Conference PaperPDF Available

Family homework activities in a comprehensive sex education program delay teen sex for boys


Abstract and Figures

Background: Communication between parents and teens about sexuality can reduce early sexual behavior, although its influence varies by gender. Get Real, a 3-year, comprehensive middle school sex education program, includes family activities with each lesson. Purpose: To describe the role of a family activity component of a sex education program in delaying first sex over the course of middle school. Significance: These findings have implications for sexual health prevention and intervention programs. Methodology: Twelve middle schools received three years of a comprehensive middle school sex education program. Baseline data was collected at the start of 6th grade in 2008-09. Follow-up data was collected at the end of 8th grade. Students' family activity participation score was calculated out of 24 possible assignments, and assessed as a dosage indicator. Logistic regression analysis controlled for the clustered study design by classroom, demographic variables, social desirability, student-reported grades, and closeness with parents. Results: At baseline the sample (N=854) was 54% female. Mean age was 11.83 years (SD=.60), and racial/ethnic background was 32% White, 30% Black, 29% Latino, 4% Asian, and 5% Biracial. After exposure to the 6th-8th grade curriculum, which included family activities with each lesson, boys who completed more family activities reported lower sexual initiation rates (OR=.96, p=.001), although there was no significant effect of family activity completion for girls. Conclusions: Participation in family activities as part of a school-based sex education program was a key input for delayed sexual initiation for middle school boys.
Content may be subject to copyright.
Protective Effects of Middle School
Comprehensive Sex Education With Family
BACKGROUND: School-based comprehensive sex education programs can reduce early adolescents’ risky sexual behavior.
The purpose of this study was to assess the effectiveness of a 3-year comprehensive sex education program in delaying vaginal
sex for middle school students and whether the family component of the intervention contributes to its effectiveness.
METHODS: This longitudinal evaluation followed a cohort of 6th graders (N =2453) through the end of 8th grade. The design
used random assignment of 24 schools into treatment and comparison conditions. The analysis included multiple-group
logistic regression to assess differences in delay of sex between intervention and comparison groups.
RESULTS: In schools where the program was taught, 16% fewer boys and 15% fewer girls had had sex by the end of 8th grade
compared to boys and girls at comparison schools. Completing family activities during the first year of the program predicted
delayed sexual debut for boys.
CONCLUSIONS: Theory-based, developmentally appropriate, comprehensive sex education programs that include parent
involvement can be effective in delaying vaginal sex for middle school students. Parent involvement is particularly important for
boys, as family activities may encourage parents to talk with their sons earlier and more frequently.
Keywords: sexuality education; adolescents; middle school; family sexuality communication; sexual behavior.
Citation: Grossman JM, Tracy AJ, Charmaraman L, Ceder I, Erkut S. Protective effects of middle school comprehensive sex
education with family involvement. J Sch Health. 2014; 84: 739-747.
Received on August 13, 2013
Accepted on April 3, 2014
Risky sexual behavior has adverse health and aca-
demic consequences for adolescents and young
adults, such as sexually transmitted infections (STIs),1,2
unintended pregnancy,3and school dropout.4,5 Evi-
dence shows that school-based comprehensive sex
education programs are associated with reductions in
adolescents’ risky sexual behavior.6,7 Some programs
also offer opportunities to engage parents in sexual-
ity communication with their teens.8Middle school
programs can play an important protective role, as
beginning sex education before teens have sex is criti-
cal in effectively reducing risky sexual behavior.9,10 In
this report we examine whether a 3-year comprehen-
sive sex education program for middle school students
that includes family activities is associated with delay-
ing vaginal sex, and whether the family component
contributes to its effectiveness.
aResearch Scientist, (jgrossma@wellesley. edu), Cheever House/Wellesley Centers for Women, Wellesley Coll ege, 106 Central Street, Wellesley, MA 02481.
bSenior Research Scienti st and Methodologist, (, Cheever House/Wellesley Centers for Women, Wellesley College, 106 Central Street, Wellesley,
MA 02481.
cResearch Scientist, (, Cheever House/Wellesley Centers for Women, Wellesley College, 106 Central Street, Wellesley, MA 02481.
School-Based Sex Education Programs
Middle school sex education programs show mixed
findings for their effectiveness in delaying sex for early
adolescents. Two evaluations of the It’s Your Game: Keep
It Real curriculum demonstrate its effect of delayed sex
for both boys and girls,11,12 whereas another program
showed no effects on sexual behavior.13 Other
research has shown that program effects can vary for
boys and girls.6Specifically, 2 randomized controlled
trial evaluations showed reduced sex for boys, but
not for girls.14,15 In contrast, another longitudinal
program evaluation found delayed sex only for
Family Components of Sex Education Programs
Evaluations seldom assess the unique contribu-
tion of family activities to the overall effectiveness
Journal of School Health November 2014, Vol. 84, No. 11 ©2014, American School Health Association 739
of school-based sex education interventions, although
adolescents name their parents as central resources in
their sexual decision-making.17 Family sexuality com-
munication can reduce teen sexual risk-taking,18-20
but parents often report lacking knowledge, skills, and
confidence to conduct those conversations.21 Programs
can support parent comfort and competence by provid-
ing specific directions and structure for communication
and resources to guide participation.8A small number
of evaluations were done on programs that include
family components and show their effectiveness in
delaying sex.12,22 In one of the few evaluations that
directly compared school-based interventions with
and without parent components, a program that did
not include parents was equally effective in reducing
frequency of recent intercourse as one that did.15
The Intervention
Get Real: Comprehensive Sex Education That Works is
a middle school sex education program developed by
the Planned Parenthood League of Massachusetts that
aims to delay sexual intercourse. As a comprehensive
program it emphasizes delaying sex as a healthy choice
while providing medically accurate information about
protection. It includes 9 lessons each in Grades 6, 7, and
8 and provides culturally sensitive and age-appropriate
information, focusing on relational skill-building as
a means to make healthy choices regarding sexual
relationships.23 The curriculum is organized around an
ecological systems perspective24 and draws from the
theory of planned behavior.25,26 Preliminary research
on the impact of the first 9 lessons27 and accompanying
family activities18 showed that both are associated with
lower rates of early sexual debut (Erkut et al27 contains
Get Real designates parents as the primary sexuality
educators of their children: the school-based cur-
riculum provides knowledge and skill-building, while
family activities (8 in each grade) give parents (or other
caring adults) an opportunity to transmit their values
about sex and relationships. Furthermore, family activ-
ities aim to increase families’ comfort in talking about
those topics (Grossman et al18 contains more details).
Family activities include talking about healthy and
unhealthy relationships; discussing media images of
sexuality; answering true/false questions about HIV
dResearch Associate, (iceder@wellesl, Cheever House/Wellesley Centers for Women, Wellesley College, 106 Central Str eet, Wellesley, MA 02481.
eSenior Research Scientist, (, Cheever House/Wellesley Centers for Women, Wellesley College, 106 Central Street, Wellesley, MA 02481.
Address correspondence to: Jennifer M. Grossman, Research Scientist, (jgrossma@wellesley. edu), Cheever House/Wellesley Centers for Women, Wellesley Coll ege, 106 Central
Street, Wellesley, MA 02481.
This res earch was supported by the Planned Parenthood League of Massachusetts for the i mpact evaluation of their comprehensi ve middle school sex education curr iculum. The
authors thank the participating schools for giving access, the students and their parents and guardians for participating, and the PPLMsexual health educators for administering
the intervention. They thank the Welles ley Centers for Women writing group f or their feedbac k and support. They would al so like to t hank Doug Kirby for consultation onresearch
design, Vincent Guilamo-Ramos and JimJaccard for their valuable feedback onthis manuscript, Alice Fryefor her analyticinput during the early phases of the research, and Erica
Plunkett for field coor dination. The research was guided by an Advisory Board composed of Barbar a Huscher Cohen, Willa Doswell, Barbar a Goodson, Jennifer Manlove, Velma
McBride Murry, Anne Noonan, Freya Sonenstei n, and Deborah Tolman, whose input is gratefully acknowledged.
and AIDS; practicing how to say ‘‘no’’ to unwanted
activities. To support parent participation in activi-
ties and overall sexuality communication with their
teens, each activity is communicated through a letter
sent home and made available online, accompanied by
The current study has 2 aims. First, it compares
treatment and comparison groups to assess the 3-year,
longitudinal impact of Get Real on delaying sex among
middle school girls and boys. Second, it examines
whether there is a unique contribution of family
activity participation on delaying sex among treatment
A total of 2453 students participated over 3 years.
Baseline surveys were administered to 2018 students
(48% treatment). In Grade 7, 1943 students took the
survey, and 1754 in Grade 8. Overall, 56% of students
completed surveys in all 3 years, while 44% either: (1)
completed the baseline survey but had a missing survey
in Grades 7 or 8; or (2) did not complete the baseline
survey, but took the 7th and/or 8th grade survey. The
fluctuating sample reflects high residential mobility
in participating school districts. Mobility tends to be
higher among students in low income households28
as is the case in our sample. Rates of residential
mobility were exacerbated by economic conditions
during and immediately following the Great Recession
(2008-2010), which also disproportionately affected
minority and low-income individuals29 and which
coincided with the beginning of data collection. We
describe in the Data Analysis section how we address
missing data resulting from students’ mobility.
Sexual activity. Students answered (yes/no) ‘‘Have
you ever had sex? Having sex means when a boy puts
his penis inside a girl’s vagina. Some people call this
‘making love’ or ‘doing it.’’’
Age. Age was calculated using date of birth.
Race/ethnicity. Students who chose any single
category for race/ethnicity were identified as such
740 Journal of School Health November 2014, Vol. 84, No. 11 ©2014, American School Health Association
(ie, Black, White, Asian, Latino). Those who reported
being biracial or chose more than 1 race were coded
as biracial/multiracial. Consistent with US Census
definitions that Hispanics can be of any race,30 students
who self-identified as any Latino ethnicity were coded
as Latino.
Two-parent family. Students were asked ‘‘Who do
you live with?’’ Responses included ‘‘2 parents in 1
place,’’ ‘‘1 parent,’’ ‘‘2 parents in different places,’’
‘‘grandparents or other family members,’’ or ‘‘other.’’
Students who chose the first were coded as having a
2-parent family structure.
Median household income. Family income level
was obtained for the census tract associated with
the address provided in the 8th grade survey. When
an address was not available, the median household
income for the tract for their school was used as a
proxy. This variable was somewhat positively skewed,
reflecting most students’ modest family incomes.
Grades. Typical grades received correspond to
response categories 1 =Mostly As,2=Mostly Asand
Bs,3=Mostly Bs,4=Mostly Bsand Cs,5=Mostly Cs,
6=Mostly Csand Ds,7=Mostly Ds,8=Mostly Dsand
Parent/guardian closeness. Combined responses
from ‘‘How close do you feel to your mother or female
guardian?’’ and ‘‘How close do you feel to your father
or male guardian?’’ (1 =notatallto5=very much)
were taken from the Grade 8 survey.
Social desirability. An abbreviated version of Bax-
ter et al31 social desirability index was used. Item scores
were summed, with higher scores representing higher
social desirability.
Get Real dosage. Student attendance at Get Real
lessons ranged from 0 to 27 possible lessons.
Get Real family activity participation. Completion
of family activities was calculated from 24 possible
This evaluation involved random assignment at the
school level into treatment and comparison conditions
of 24 middle schools in the greater Boston area (13
public, 9 charter, and 2 private). Schools signed up
before random assignment, which avoids bias due to a
school’s preference for either condition. In treatment
schools, educators trained by the curriculum developer
taught all classes, whereas students in comparison
schools received sex education as currently taught in
their schools. To assess change in their knowledge,
attitudes, and behaviors related to relationships and
sexual health, students were administered surveys at
3 time points: beginning of Grade 6 (baseline, starting
in 2009), beginning of Grade 7, and end of Grade 8
(completed by 2012).
Parent/guardian informed consent forms were
distributed in multiple languages. Schools decided
whether to use passive or active consent. All 12
treatment schools chose passive consent whereas 5
comparison schools chose active consent. Eighteen
parents in active consent schools opted out, and 84 did
not return consent forms in 6th grade (30% of students
in those 5 schools; 4% of total students). Thirteen
parents opted out in passive consent schools (<1%
of students in those schools; <1% of all students).
By the time students were in 8th grade, all schools
had adopted passive consent. Although active consent
schools had a higher proportion of white students and
students from 2-parent families, student race/ethnicity
and family structure did not predict missing data. All
students provided written assent prior to completing
surveys, with an average of 9 students opting out
each year.
Data Analysis
We conducted analyses separately for girls and
boys for several reasons. First, prior sex education
evaluations have shown mixed effects on girls’ and
boys’ sexual behavior.14,16 Second, boys’ and girls’
sexual experiences and processes may differ due to
sex-specific physiological changes of adolescence32 and
proscribed norms for dating and sexual activity.33,34
Finally, baseline rates of self-reported sexual activity
differed significantly by sex; 3.6% of girls, compared
to 10.5% of boys, reported having had sex by the
start of Grade 6 (χ2=30.30, df =1, p <.001, Odds
ratio =0.32). By conducting the analyses separately,
we also avoid the assumption that the covariates
have the same effects on having had sex regardless
of gender.
A number of control variables were included
as covariates in our regression models. Older age
and lower levels of family income are known risk
factors for early sexual activity, whereas living with
2 parents is protective.35 These characteristics, as well
as race, differed significantly between treatment and
comparison groups at baseline (Table 1) and were,
therefore, included as covariates. Social desirability
and students’ grades were included to distinguish their
influences from program effects. We also included 6th
grade reports of sexual activity because 6th graders
may not have understood and defined sexual behavior
accurately,36 and therefore, may have incorrectly
reported having had sex. As is standard practice when
comparing groups, we centered covariates on their
grand mean, rather than on group-specific means and
held covariate effects to be invariant across groups so
that adjusted outcomes are directly comparable across
Typically, tests of treatment effects are conducted
using a simple logistic regression equation model
Journal of School Health November2014,Vol.84,No.11 ©2014, American School Health Association 741
Table 1. Descriptive Characteristics of the Comparison and Treatment Groups
Comparison (N =1231) Treatment (N =1221)
N Mean SD N Mean SD Cohen’s d t-test df
Age at 8th 1231 14.00 0.74 1221 13.88 0. 78 .16 4.05 2450***
Median household income 653 $ 46,416 20266 684 $ 57,964 27646 .48 8.68 1335***
Social desirability bias 1178 0.31 0.27 1185 0.28 0.27 .11 2.46 2361*
Poor grades at baseline 642 2.90 1.45 810 2.76 1.38 .10 1. 90 1450
Dosage during 6th 1221 6.71 3.36 n/a
Dosage during 7th and 8t h 1221 10.17 6. 73 n/a
Activities during 6th 1221 2.72 2.97 n/a
Activit ies during 7th and 8th 1221 3.74 4.49 n/a
N % N % Odds Ratio χ2(1)
Female 1231 48% 1221 55% 1.32 12.12 ***
Biracial 1231 14% 1221 14% 1. 00 0.10
Latino 1231 37% 1221 35% .92 1.02
Black 1231 42% 1221 32% .65 36.12 ***
White 1231 14% 1221 31% 2.76 111.38 ***
Asian 1231 7% 1221 4% .55 15.51 ***
2-parent household 771 47% 789 51% 1.17 6.71 *
Baseline sexual activity 757 6% 868 7% 1.18 0.56
8th grade sexual activity 799 28% 877 21% .68 10.92 **
*p <.05, **p <.01, ***p <.001.
Note: For dosage and family activities, Grades 7 and 8 were combined because assessments took place at the beginning of Grade 7 and at the end of Grade 8, a 2-year span.
in which treatment assignment, along with relevant
covariates, predicts sexual outcomes. However, the
clustered sampling design, missing data due to
residential mobility leading to attrition, and our
need to assess both overall group effects and
effects of family activities within the treatment
group introduce complexities better handled by a
more flexible modeling paradigm, namely structural
equation modeling (SEM).37,38 We apply a multiple
group logistic regression model with adjustment for
the clustered sampling design and with missing data
handling, using the Mplus statistical modeling program
Clustered Sampling Design
Regression modeling assumes that the sample is
comprised of independent cases. However, when
students are sampled from within schools, their
responses may not be truly independent.40 Therefore,
similarity of students within schools must be accounted
for in the analysis model. As Asparouhov and
en41 advocate for sampling designs with few
clusters, we used an adjusted LRT statistic to account
for this design effect.42
To assess both overall group effects and the
influence of family activities within the treatment
group, we estimated group differences in likelihood
of sexual onset in a multiple group model in which
covariate effects are held invariant across groups. As
en and Satorra43 explain, the intercept of a
logistic regression equation estimated as a structural
equation model is expressed as a threshold parameter.
A comparison of thresholds estimated for the treatment
and comparison group constitutes a test of the relative
risk across groups.44
Missing Data Handling
As noted in the Participants section, our sample
reflected a highly mobile population. We have 2
groups of ‘‘movers,’’ namely ‘‘leavers’’ who contribute
only baseline data and subsequently leave the study,
and ‘‘joiners’’ who contribute only follow-up data
by joining after baseline. Without including these
students, we run the risk of biasing our results to reflect
only the ‘‘stayers’’those who remain across the
entire period, which would restrict the generalizability
of the findings.45 Differences between ‘‘movers’’ and
‘‘stayers’’ are a real concern in our sample. ‘‘Movers’’
reported higher likelihood of having had sex at baseline
(t =7.75, p <.01) and at the end of Grade 8 (t =10.17,
p<.01). They reported demographic characteristics
that have been associated with higher sexual risk,7,35
including lower median income (t =4.17, p <.01) and
lower likelihood of living with 2 parents (t =8.70,
p<.01). There were also racial differences (χ2=80.02,
p<.001), with ‘‘movers’’ more likely to be black and
less likely to be white. By including ‘‘movers’’ in the
analyses we were able to retain students who have
higher risk for early sexual debut.
We also assessed differences between the 2 groups
of ‘‘movers.’’ We did not find significant differences
with respect to age, sex, median household income,
family structure, race/ethnicity (with the exception of
the biracial category), or the likelihood of having had
742 Journal of School Health November2014,Vol.84,No.11 ©2014, American School Health Association
sex. This comparability supports the argument that
data from participants who joined after baseline can
be used to offset attrition to more accurately estimate
treatment effects of students in participating schools.
For these reasons, we used strategies to statistically
handle the resulting missing data rather than conduct-
ing analyses with listwise deletion.46 Estimated values
for missing independent variables were imputed (Soul-
lier et al47 describe this approach) using the NORM sta-
tistical program, which generates empirically derived
values for missing data in continuous and categorical
variables.48 In contrast, missingness in the dependent
variable was handled by means of full-information
maximum likelihood estimation (FIML), which adjusts
standard errors of the estimates rather than assigning
individual values. Based on the assumption that the
data are missing at random (MAR) or missing com-
pletely at random (MCAR), these methods of handling
missing data are unbiased.49 We conducted a set of post
hoc analyses (described briefly in the Results section)
to test the robustness of our results with respect to
violations of this assumption.
Effects Within the Treatment Group
A second set of models estimated variability in risk
within the treatment group by examining predictive
effects of Get Real lesson attendance (dosage) and
completion of family activities. This poses a challenge
in a multiple group model because dosage and
activities do not exist in the comparison group. To test
differences within and across groups simultaneously,
we used ‘‘phantom variables’’ to estimate these
models.50 We introduced a small amount of random
noise with a negligible range (.001 to .001) in place
of dosage and family activities in the comparison
group and set to zero the variances and covariances
of these variables and their corresponding regression
effects, while freely estimating these parameters in the
treatment group. For these models, all the previous
covariates were included along with parent/guardian
closeness, an important control when exploring the
effects of family activities, to assess whether the
influence of family communication about sex extends
beyond the effects of close family relationships.
Baseline Covariates in Treatment and Comparison
Treatment group participants’ mean age in 8th grade
was 13.88 (SD =.78), younger than the mean age of
14.00 in the comparison group (SD =.74); they were
also more likely to be white (31% vs. 14%), whereas
comparison group participants were more likely to be
black (42% vs. 32%) or Asian (7% vs. 4%). More
treatment participants lived with 2 parents in the
same home (51% vs. 47%) and had a higher median
family income ($57,964 vs. $44,416) but there were
no significant differences between groups in report of
having had sex at baseline. Table 1 contains descriptive
statistics and tests of mean differences between groups.
Intervention and Comparison Group Outcomes
Tables 2 and 3 present results of the models testing
treatment/comparison group differences for girls and
boys respectively, adjusted for covariate effects, where
a higher threshold value implies a lower probability of
having had sex. Table 2 shows that, for girls, there is a
significant difference between the groups (thresholds:
comparison =1.03, treatment =1.24, Wald test =3.98,
df =1, p <.05). The adjusted rate of sexual debut
(or the implied likelihood of sexual debut) for girls
in the treatment group, 22.4%, is 15% lower than
the adjusted rate for girls in the comparison group,
26.3%. Similarly, Table 3 shows that, for boys, the
difference between the treatment and comparison
group in the likelihood of having had sex by Grade
8 is also significant (thresholds: comparison =0.42,
treatment =0.70, Wald test =4.04, df =1, p <.05).
For boys, the adjusted rate of sexual debut for the
treatment group, 33.2%, is 16% lower than the
adjusted rate for the comparison group, 39.6%.
Treatment Effects Within the Intervention Group
In the models testing both between-group differ-
ences and within-group effects neither dosage nor
family activity effects are significant for girls, but the
effect of completing family activities during Grade
6 was statistically significant for boys (B =−0.04,
SE =0.02, odds ratio =0.97, 95% CI =0.934,0.998;
Table 4). In these models, the overall effect of partic-
ipating in Get Real remains statistically significant for
both girls and boys.
Sensitivity Analysis
We tested the robustness of our results to missing
data by conducting sensitivity analyses and comparing
results of analyses with missing data handling in
the full sample to results obtained with a smaller
sample of only students who provided complete data
(listwise deletion). With the exception of the model
for girls, from which baseline reports of having had
sex were excluded as they were too rare to model,
these analyses used parallel models. Similar to our
models with missing data handling, analyses that
used listwise deletion showed that treatment students
were significantly less likely to have had sex (girls:
thresholds: comparison =1.03, treatment =1.62,
Wald test =5.03, df =1, p <.05; boys: thresholds:
comparison =.40, treatment =1.26, Wald test =9.88,
df =1, p <.01). The results of this sensitivity analysis
support the robustness of our findings.
Journal of School Health November 2014, Vol. 84, No. 11 ©2014, American School Health Association 743
Table 2. Results of the Multiple Group Logistic Regression Model Predicting the Probability of Becoming Sexually Active by the
End of 8th Grade, Controlling for Clustering at the School Level—Girls
Comparison (N =592) Treatment (N =673) Test of Group Differences
Estimate SE Estimate SE Wald (1)
Threshold 1.03 0.07 1.24 0.08 3.98*
Implied Proportion 26.3% 22.4%
B SE OR (95% CI) B SE OR (95% CI)
Baseline sexual activit y 1.35 0.18 3. 85 (2.719-5.462) 1.35 0.18 3.85 (2.719-5.462)
Age at 8th 0. 23 0.04 1. 26 (1.159-1.372) 0.23 0.04 1.26 (1. 159-1.372)
Biracial 0.83 0.21 2.28 (1.524-3.417) 0.83 0.21 2.28 (1.524-3.417)
Latino 0.59 0. 16 1.81 (1.312-2.486) 0.59 0.16 1.81 (1. 312-2.486)
Black 0.45 0. 16 1.57 (1.135-2.159) 0.45 0.16 1.57 (1. 135-2.159)
2-parent househol d 0.38 0. 11 0.68 (0.554-0.840) 0.38 0.11 0.68 (0.554-0.840)
Median household i ncome 0.12 0.07 0.89 ( 0.778-1.019) 0. 12 0.07 0.89 (0.778-1.019)
Poor grades at baseline 0.17 0.07 1.18 (1.032-1.358) 0.17 0.07 1.18 (1.032-1. 358)
Social desirabili ty bias 0.13 0. 05 0.88 (0.793-0.972) 0.13 0.05 0.88 (0.793-0.972)
*p <.05, **p <.01, ***p <.001.
Note: Covariate effects have been constrained to be equal across groups to reflect direct effects on having had sex, rather than interaction effects with treatment group
Table 3. Results of the Multiple Group Logistic Regression Model Predicting the Probability of Becoming Sexually Active by the
End of 8th Grade, Controlling for Clustering at the School Level—Boys
Comparison (N =639) Treatment (N =548) Test of Group Differences
Estimate SE Estimate SE Wald (1)
Threshold 0.42 0.11 0.70 0.09 4.04*
Implied Proportion 39.6% 33.2%
B SE OR (95% CI) B SE OR (95% CI)
Baseline sexual activit y 1.14 0.16 3. 13 (2.290-4.270) 1.14 0.16 3.13 (2.290-4.270)
Age at 8th 0. 24 0.04 1. 27 (1.182-1.372) 0.24 0.04 1.27 (1. 182-1.372)
Biracial 1.03 0.18 2.79 (1.943-3.997) 1.03 0.18 2.79 (1.943-3.997)
Latino 1.17 0. 14 3.23 (2.468-4.223) 1.17 0.14 3.23 (2. 468-4.223)
Black 1.06 0. 17 2.88 (2.087-3.984) 1.06 0.17 2.88 (2. 087-3.984)
2-parent househol d 0.38 0. 08 0.69 (0.586-0.802) 0.38 0.08 0.69 (0.586-0.802)
Median household i ncome 0.12 0.05 0.89 ( 0.806-0.980) 0. 12 0.05 0.89 (0.806-0.980)
Poor grades at baseline 0.01 0. 05 0.99 (0.905-1.088) 0.01 0.05 0.99 (0.905-1.088)
Social desirabili ty bias 0.07 0. 04 0.94 (0.862-1.016) 0.07 0.04 0.94 (0.862-1.016)
p<.05, **p <.01, ***p <.001.
Note: Covariate effects have been constrained to be equal across groups to reflect direct effects on having had sex, rather than interaction effects with treatment group
Boys and girls both benefit from Get Real, making
it one of only a few middle school programs that
show reduced risky sexual behavior for both boys
and girls.11,12 This suggests that a theory-based
program that provides developmentally appropriate
information and builds skills to negotiate healthy
relationships can delay sexual debut for middle school
students. Our finding that the completion of family
activities during the first year of the program predicted
delayed sexual debut for boys, suggests that early
support for family communication was particularly
critical for boys’ sexual health.
In schools where Get Real was taught 16%
fewer boys and 15% fewer girls had had sex by
Grade 8 compared to students who received sex
education curriculum as currently taught in their
schools. This program’s success may reflect both its
focus on relational skill-building to help adolescents
translate their intentions into behaviors, as well as
its developmental approach, by targeting lessons to
students’ grade and maturity level.
The additional delay in having sex among boys
who complete 6th grade family activities extends prior
findings for effectiveness of sex education programs
that include family components, but that do not
directly assess their contribution to sexual behavior
outcomes.12,22 Including family activities in Grade 6
may be influencing boys’ behavior by encouraging
parents and sons to talk about sexual issues earlier
744 Journal of School Health November 2014, Vol. 84, No. 11 ©2014, American School Health Association
Table 4. Results of the Multiple Group Logistic Regression Model Predicting the Probability of Becoming Sexually Active by the
End of 8th Grade, Controlling for Clustering at the School Level—Boys
Comparison (N =639) Treatment (N =548) Test of Group Differences
Estimate SE Estimate SE Wald (1)
Threshold 0.42 0.12 1.23 0.32 5.55*
B SE OR (95% CI) B SE OR (95% CI)
Baseline sexual activity 1.00 0.15 2.72 (2.026-3.647) 1.00 0.15 2.72 (2.026-3.647)
Age at grade 8 0.26 0.05 1.29 (1.176-1.425) 0.26 0. 05 1.29 (1.176-1.425)
Biracial 0.98 0.18 2.66 (1.867-3.795) 0.98 0.18 2.66 (1.867-3.795)
Latino 1.05 0.13 2.85 (2.193-3.709) 1.05 0.13 2.85 (2.193-3.709)
Black 0.92 0.15 2.51 (1.885-3.341) 0.92 0.15 2.51 (1.885-3.341)
Two-parent household 0.31 0.08 0.73 (0.632-0.851) 0.31 0.08 0.73 (0.632-0.851)
Medi an HH income 0.12 0.05 0. 89 (0.813-0.977) 0.12 0.05 0.89 (0.813-0.977)
Poor grades at baseline 0.01 0.05 1.01 (0.923-1.105) 0.01 0.05 1.01 (0.923-1.105)
Parent closeness 0.09 0.06 0.91 (0.812-1. 027) 0.09 0. 06 0.91 (0.812-1.027)
Social desirabili ty bias 0.21 0.14 0.81 (0.624-1.063) 0.21 0.14 0.81 (0.624-1.063)
Dosage during 6th 0 0 0 0.01 0.01 1.01 (0.983-1. 034)
Dosage during 7th and 8th 0 0 0 0.01 0.02 1.01 (0. 977-1.040)
Activities during 6th 0 0 0 0. 04 0.02 0.97 (0.934-0.998)
Activities during 7th and 8th 0 0 0 0.03 0.03 0.97 ( 0.910-1.031)
p<.05, **p <.01, ***p <.001.
Note: Covariate effects have been constrained to be equal across groups to reflect direct effects on having had sex, rather than interaction effects with treatment group
assignment. Dosage and activities have been centered at the maximum value (best case scenario).
than they would have otherwise, which is critical to
delaying sex.9,10 Moreover, they also may increase
the frequency of these conversations, which is often
greater for girls than boys.51,52 Therefore, Get Real
may promote change in both the starting point and
frequency of conversations about sex between boys
and their families. Findings for the effectiveness of
family activities for boys but not for girls also suggest
the need for further research that goes beyond the
‘‘does it work’’ question, to investigate ‘‘which part’’
and ‘‘for whom.’’
We did not find a significant effect for dosage
for boys or girls. This was somewhat surprising, as
we would expect that students who attended more
lessons would show reduced risk. However, the lack
of statistical significance may reflect how dosage was
modeled, namely estimating the effect of each lesson
attended. The small effect of a single lesson on delayed
sexual debut that is in the expected direction, albeit
not statistically significant, suggests that there may be
a protective cumulative influence over the 3 years.
These results are informative in light of nega-
tive educational and health correlates of early sexual
debut, like school dropout,4,5 sexually transmitted
infections,1,2 and unintended pregnancy.3Nation-
wide, 18% of adolescents report having had sex by
Grade 8 (30% of 8th graders in urban schools).53
Given findings that communication about sexual issues
should begin before teens become sexually active,9,10
programs that support early access to both school- and
parent-based sex education are important to adoles-
cents’ healthy development.
The high level of participant mobility is considered
a critical threat to the generalizability of an impact
evaluation, due to the likelihood of sampling bias. Our
strategy of including data from students who enter
the study as a ‘‘replacement’’ for lost information
from students who leave is our attempt to avoid
sampling bias due to mobility, which affects most
long-term impact evaluations of interventions with
at-risk populations. The defensibility of this strategy
rests on the comparability of ‘‘leavers’’ and ‘‘joiners’’
which is supported for the demographics we tested
in this study, and on the viability of the missing
data handling employed in the analysis. Moreover,
violations to MAR/MCAR assumptions for missing
data, even in conditions of high attrition, often have
minimal impact on results of impact evaluations.46
Baseline differences between treatment and com-
parison groups are also a limitation. The goal of
random assignment is to create equal groups. How-
ever, working with a relatively low number of cases
(such as 24 schools) can result in the groups being
imbalanced by chance. Although we attempted to
address this concern by including variables in which
these groups differed as controls in study analyses,
we cannot dismiss the possibility that the groups
represented populations of students with different
characteristics not captured in our models. Therefore,
we encourage larger-scale replication studies to test the
generalizability of our results.
Although we included a social desirability measure
to buffer the limitations of self-report data, we
Journal of School Health November 2014, Vol. 84, No. 11 ©2014, American School Health Association 745
acknowledge the need for more diverse data collection
methods. The focus on vaginal sex limits applicability
to students who engage in non-vaginal sex, and does
not fully assess sexual behaviors of lesbian, gay, and
bisexual students. Whereas other types of sex were
addressed within the Get Real curriculum, the impact
of the program on these behaviors was not assessed
through this evaluation.
In sum, findings support the assertion that theory-
based, developmentally appropriate, comprehensive
sex education programs that include parent involve-
ment can be effective in delaying sex for middle
school students. Parent involvement may be partic-
ularly important for boys, as family activities may
encourage parents to talk with their sons earlier
and more frequently. These results support other
findings that early sex education interventions are
important for protecting youth from the negative
health and academic consequences of early sexual
Despite research documenting the effectiveness of
comprehensive sex education programs,12,14 state poli-
cies vary in their requirements for sex education.54
In addition, research shows high levels of par-
ent support for comprehensive sex education,
which is at times inconsistent with state prac-
tice and federal funding for abstinence-only sex
education.55,56 Greater connection between research,
policy, and practice can help to support programs with
actual potential to reduce adolescents’ risky sexual
Get Real provided online and paper resources to
support parents’ sexuality communication with their
teens. Parents differ in their receptiveness to talking
with their middle school-aged children about sexual
health. In some locales more outreach and education
may be needed to further develop and strengthen
parent components of sex education programs. Once
programs are in place, support is needed to maximize
student attendance and completion of family activities
in order to achieve full protective effects. School
educators and counselors can encourage parents and
teens to talk to each other about sex by linking them
to critical resources.57,58
Human Subjects Approval Statement
The research was approved by Wellesley College’s
Institutional Review Board and by relevant review
bodies for participating schools. The study adhered to
all human subjects’ protections.
1. Kaestle CE, Halpern CT, Miller W, Ford CA. Young age at
first sexual intercourse and sexually transmitted infections
in adolescents and young adults. Am J Epidemiol. 2005;161:
2. Kahn JA, Rosenthal SL, Succop PA, Ho GYF, Burk RD.
Mediators of the association between age of first sexual
intercourse and subsequent human papillomavirus infection.
Pediatrics. 2002;109:e5.
3. von Ranson K, Rosenthal S, Biro F, Lewis L, Succop P.
Longitudinal risk of STD acquisition in adolescent girls using a
generalized estimating equations model. J Pediatr Adolesc Gynecol.
4. Frisco ML. Adolescents’ sexual behavior and academic
attainment. Sociol Educ. 2008;81(3):284-311.
5. Steward NR, Farkas G, Bingenheimer JB. Detailed educational
pathways among females after very early sexual intercourse.
Perspect Sex Reproduct Health. 2009;41(4):244-252.
6. Chin HB, Sipe TA, Elder R, et al. The effectiveness of
group-based comprehensive risk-reduction and abstinence
education interventions to prevent or reduce the risk of
adolescent pregnancy, human immunodeficiency virus, and
sexually transmitted infections: two systematic reviews for
the Guide to Community Preventive Services. Am J Prev Med.
7. Kirby D. Emerging Answers 2007: Research Findings on Programs
to Reduce Teen Pregnancy and Sexually Transmitted Diseases.
Washington, DC: National Campaign to Prevent Teen and
Unplanned Pregnancy; 2007. Available at:
ERICWebPortal/contentdelivery/servlet/ERICServlet?accno =
ED456171. Accessed January 10, 2012.
8. Kirby D, Miller BC. Interventions designed to promote parent-
teen communication about sexuality. In: Feldman S, Rosenthal
D, eds. Talking Sexuality: Parent-Adolescent Communication.San
Francisco, CA: Jossey-Bass; 2002:93-110.
9. Clawson CL, Reese-Weber M. The amount and timing
of parent-adolescent sexual communication as predictors
of late adolescent sexual risk-taking behaviors. JSexRes.
10. Miller K, Levin ML, Whitaker DJ, Xu X. Patterns of condom
use among adolescents: the impact of maternal-adolescent
communication. Am J Public Health. 1998;88:1542-1544.
11. Markham CM, Tortolero SR, Peskin MF, et al. Sexual risk
avoidance and sexual risk reduction interventions for middle
school youth: a randomized controlled trial. J Adolesc Health.
12. Tortolero SR, Markham CM, Peskin MF, et al. It’s Your Game:
Keep It Real: delaying sexual behavior with an effective middle
school program. J Adolesc Health. 2010;46(2):169-179.
13. Koo HP, Rose A, El-Khorazaty MN, et al. Evaluation of a
randomized intervention to delay sexual initiation among fifth
graders followed through the sixth grade. Sex Educ. 2011;
14. Coyle KK, Kirby DB, Marin BV, Gomez CA, Gregorich SE. Draw
the Line/Respect the Line: a randomized trial of a middle school
intervention to reduce sexual risk behaviors. Am J Public Health.
15. Flay BR, Graumlich S, Segawa E, Burns JL, Holliday MY. Aban
Aya Investigators. Effects of 2 prevention programs on high-risk
behaviors among African American youth. Arch Pediatr Adolesc
Med. 2004;158(4):377-384.
16. Aarons SJ, Jenkins RR, Raine TR. El-Khorazaty. Postpon-
ing sexual intercourse among urban junior high school
students— a randomized controlled evaluation. J Adolesc Health.
17. Albert B. With One Voice 2010: America’s Adults and Teens Sound off
about Teen Pregnancy. Washington, DC: The National Campaign
to Prevent Teen and Unplanned Pregnancy; 2010.
746 Journal of School Health November2014,Vol.84,No.11 ©2014, American School Health Association
18. Grossman JM, Frye A, Charmaraman L, Erkut S. Family
homework and school-based sex education: delaying early
adolescents’ sexual behavior. J Sch Health. 2013;83(11):
19. Trejos-Castillo E, Vazsonyi AT. Risky sexual behaviors in first
and second generation Hispanic immigrant youth. J Youth
Adolesc. 2009;38(5):719-731.
20. Usher-Seriki KK, Bynum MS, Callands TA. Mother-daughter
communication about sex and sexual intercourse among
middle- to upper-class African American girls. J Fam Issues.
21. Guilamo-Ramos V, Jaccard J, Dittus P, Collins S. Parent-
adolescent communication about sexual intercourse: an analysis
of maternal reluctance to communicate. Health Psychol.
22. O’Leary A, Jemmott JB III, Jemmott LS, et al. Moderation and
mediation of an effective HIV risk-reduction intervention for
South African adolescents. Ann Behav Med. 2012;44(2):181-191.
23. Planned Parenthood League of Massachusetts. Empowering
youth: how Get Real works. Available at: http://www.getreal Accessed June
4, 2013.
24. Bronfenbrenner U. Contexts of child rearing: problems and
prospects. Am Psychol. 1979;34(10):844-850.
25. Ajzen I. The theory of planned behavior. Organ Behav Hum Decis
Processes. 1991;50(2):179-211.
26. Ajzen I. Behavioral interventions based on the theory of
planned behavior. Available at:
aizen/pdf/tpb.intervention.pdf. Accessed June 4, 2013.
27. Erkut S, Grossman JM, Frye A, Ceder I, Charmaraman L, Tracy
AJ. Can sex education delay early sexual debut? J Early Adolesc.
28. O’Donnell R, Gazos A. Student mobility in Massachusetts. Education
Research Brief, Office of Strategic Planning, Research, and Evaluation.
Boston, MA: Massachusetts Department of Elementary and
Secondary Education; 2010.
29. Stoll MA. Great Recession spurs a shift to local moves. Research brief .
New York, NY: The Russell Sage Foundation; 2013. Avail-
able at:
recession-spurs-shift-to-local-moves. Accessed June 4,
30. Ennis, SR, Rios-Vargas M, Albert NG. The Hispanic popu-
lation, 2010. U.S. Census. Available at: http://www.census.
gov/prod/cen2010/briefs/c2010br-04.pdf. Accessed June 4,
31. Baxter SD, Smith AF, Litaker MS, Baglio ML, Guinn CH, Shaffer
NM. Children’s social desirability and dietary reports. J Nutr Educ
Behav. 2004;36(2):84-89.
32. Steinberg L. Adolescence. 8th ed. New York, NY: McGraw-Hill;
33. Montgomery MJ. Psychosocial intimacy and identity: from
early adolescence to emerging adulthood. J Adolesc Res.
34. Savin-Williams RC, Diamond LM. Sex. In: Lerner RM, Steinberg
L, eds. Handbook of Adolescent Psychology. 2nd ed. Hoboken, NJ:
John Wiley & Sons; 2004:189-231.
35. Abma JC, Martinez GM, Mosher WD, Dawson BS. Teenagers
in the United States: sexual activity, contraceptive use, and
childbearing, 2002. Vital Health Stat 2004; 23. Available
Accessed May 2, 2013.
36. Charmaraman L, Lee AJ, Erkut S. ‘What if you already know
everything about sex?’ Content analysis of questions from early
adolescents in a middle school sex education program. J Adolesc
Health. 2012;50(5):527-530.
37. Kline RB. Principles and Practice of Structural Equation Modeling.
3rd ed. New York, NY: Guilford Press; 2010.
38. Carter RL. Solutions for missing data in structural equation
modeling. Res Pract Assess. 2006;1(1):1-6.
39. Muth´
en LK, Muth´
en BO. MplusUsersGuide.7thed.Los
Angeles, CA: Muth´
en & Muth´
en; 1998-2012.
40. Cochran WG. Sampling Techniques. New York, NY: John Wiley
& Sons; 2007.
41. Asparouhov T, Muthen B. Comparison of estimation meth-
ods for complex survey data analysis. Mplus Web Notes:
2006. Available at:
SurveyComp21.pdf. Accessed August 1, 2013.
42. Muth´
en BO, Satorra A. Technical aspects of Muth´
LISCOMP approach to estimation of latent variable relations
with a comprehensive measurement model. Psychometrika.
43. Asparouhov T, Muthen B. Multivariate statistical modeling
with survey data. Paper presented at the Federal Committee
on Statistical Methodology (FCSM) Research Conference;
November, 14-15, 2005; Arlington, VA.
44. Gl ¨
ockner-Rist A, Hoijtink H. The best of both worlds:
factor analysis of dichotomous data using item response
theory and structural equation modeling. Struct Equat Model.
45. Miller RB, Hollist CS. Attrition bias. In: Salkind NJ, ed.
Encyclopedia of Measurement and Statistics,Vol.1.ThousandOaks,
CA: Sage; 2007:57-60.
46. Graham JW, Hofer SM, Donaldson SI, MacKinnon DP,
Schafer JL. Analysis with missing data in prevention research.
In: Bryant KJ, Windle M, West SG, eds. The Science of
Prevention: Methodological Advances from Alcohol and Substance
Abuse Research. Washington, DC: American Psychological
Association; 1997:325-366.
47. Soullier N, de La Rochebrochard E, Bouyer J. Multiple
imputation for estimation of an occurrence rate in cohorts
with attrition and discrete follow-up time points: a simulation
study. BMC Med Res Methodol. 2010;10:79-85.
48. Schafer JL. Analysis of Incomplete Multivariate Data. University
Park, PA: Chapman & Hall; 1997.
49. Graham JW. Adding missing-data-relevant variables to FIML-
based structural equation models. Struct Equat Model.
50. Rindskopf D. Using phantom and imaginary latent variables
to parameterize constraints in linear structural models.
Psychometrika. 1984;39(1):37-47.
51. Omar H, McElderry D, Zakharia R. Educating adolescents
about puberty: what are we missing? Int J Adolesc Med Health.
52. Swain CR, Ackerman LK, Ackerman MA. The influence of
individual characteristics and contraceptive beliefs on parent-
teen sexual communications: a structural model. J Adolesc
Health. 2006;38(6):753.e9-753.e18.
53. Shanklin SL, Brener N, McManus T, Kinchen S, Kann L.
2005 Middle School Youth Risk Behavior Survey. Atlanta, GA: US
Department of Health and Human Services, Centers for Disease
Control and Prevention; 2007.
54. National Conference of State Legislatures. Available at:
sex-education-in-schools.aspx. Accessed June 3, 2013.
55. Barr EM, Moore MJ, Johnson T, Forrest J, Jordan M.
New evidence: data documenting parental support for earlier
sexuality education. J Sch Health. 2014;84(1):10-17.
56. Constantine NA, Jerman P, Huang AX. California parents’
preferences and beliefs regarding school-based sex education
policy. Perspect Sex Reproduct Health. 2007;39(3):167-175.
57. National Campaign to Prevent Teen and Unplanned Preg-
nancy. Available at:
parents/ten_tips.aspx. Accessed June 4, 2013.
58. Centers for Disease Control and Prevention. Division of Repro-
ductive Health, National Center for Chronic Disease Preven-
tion and Health Promotion. Available at:
teenpregnancy/Parents.htm. Accessed June 4, 2013.
Journal of School Health November 2014, Vol. 84, No. 11 ©2014, American School Health Association 747
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
In this study, we examine whether a nine-lesson sex education intervention, "Get Real: Comprehensive Sex Education That Works," implemented in sixth grade, can reduce the number of adolescents who might otherwise become "early starters" of sexual activity (defined as heterosexual intercourse) by seventh grade. Participants were 548 boys and 675 girls who completed surveys in both sixth grade (baseline) and seventh grade (follow-up). The sample was 35% Latino, 32% Black, 24% White, 3% Asian, and 6% biracial. Students randomly assigned to the control condition were 30% more likely to initiate sex by follow-up when controlling for having had sex by sixth grade, demographic variables, and a tendency to give socially desirable responses. This finding is noteworthy because previous research has identified early starters to be prone to poor outcomes in sexual health, family formation, economic security, and incarceration and few middle school interventions have shown an effect on behavioral outcomes.
Full-text available
Attrition bias is one of the major threats to multiwave studies, and it can bias the sample in two ways. First, attrition bias can affect the external validity of the study. If some groups of people drop out of the study more frequently than others, the subsequent longitudinal sample no longer resembles the original sample in the study. As a result, the remaining sample is not generalizable to the original population that was sampled. For example, a longitudinal sample examining the grieving process of women following the death of a spouse may fail to retain those participants who have become too distraught to fill out the questionnaire. The nonparticipation of this group may bias the findings of the study toward a minimization of depressive symptomatology as a component of the grieving process. In other words, the composition of the sample changes to the point that the results are no longer generalizable to the original population of widows.
Numerous studies document support for sexuality education to be taught in high school, and often, in middle school. However, little research has been conducted addressing support for sexuality education in elementary schools. As part of the state Behavioral Risk Factor Surveillance System (BRFSS) Survey administration, the Florida Department of Health conducted the Florida Child Health Survey (FCHS) by calling back parents who had children in their home and who agreed to participate (N = 1715). Most parents supported the following sexuality education topics being taught specifically in elementary school: communication skills (89%), human anatomy/reproductive information (65%), abstinence (61%), human immunodeficiency virus (HIV)/sexually transmitted infections (STIs) (53%), and gender/sexual orientation issues (52%). Support was even greater in middle school (62-91%) and high school (72-91%) for these topics and for birth control and condom education. Most parents supported comprehensive sexuality education (40.4%), followed by abstinence-plus (36.4%) and abstinence-only (23.2%). Chi-square results showed significant differences in the type of sexuality education supported by almost all parent demographic variables analyzed including sex, race, marital status, and education. Results add substantial support for age-appropriate school-based sexuality education starting at the elementary school level, the new National Sexuality Education Standards, and funding to support evidence-based abstinence-plus or comprehensive sexuality education.
Early sexual activity can undermine adolescents' future school success and health outcomes. The purpose of this study was to assess the role of a family homework component of a comprehensive sex education intervention in delaying sexual initiation for early adolescents and to explore what social and contextual factors prevent adolescents from completing these family homework activities. This mixed methods study included 6th- and 7th-grade survey responses from 706 students at 11 middle school schools receiving a sex education intervention, as well as interviews from a subset of 33, 7th-grade students from the larger sample. Adolescents who completed more family homework assignments were less likely to have vaginal intercourse in 7th grade than those who completed fewer assignments, after controlling for self-reports of having had vaginal intercourse in 6th grade and demographic variables. Participants' explanations for not completing assignments included personal, curriculum, and family-based reasons. Family homework activities designed to increase family communication about sexual issues can delay sex among early adolescents and contribute to school-based sex education programs. Successful sex education programs must identify and address barriers to family homework completion.