Dev-ined.fr

Explaining Recent Declines in Adolescent Pregnancy in the United States: The Contribution of Abstinence and Improved Contraceptive Use | John S. Santelli, MD, MPH, Laura Duberstein Lindberg, PhD, Lawrence B. Finer, PhD, and Susheela Singh, PhD In recent years, the United States has had the Objectives. We explored the relative contributions of declining sexual activity highest rate of adolescent pregnancy of any of and improved contraceptive use to the recent decline in adolescent pregnancy the world’s developed nations.1,2 However, since 1991 these rates have declined dra- Methods. We used data from 1995 and 2002 for women 15 to 19 years of age to matically. Pregnancy rates among 15- to 19- develop 2 indexes: the contraceptive risk index, summarizing the overall effec- year-olds declined 27% from 1991 to 2000,3 tiveness of contraceptive use among sexually active adolescents (including nonuse), and birth rates (for which more recent pub- and the overall pregnancy risk index, calculated according to the contraceptive riskindex score and the percentage of individuals reporting sexual activity.
lished data are available) dropped 33% be- Results. The contraceptive risk index declined 34% overall and 46% among adolescents aged 15 to 17 years. Improvements in contraceptive use included increases in the use of condoms, birth control pills, withdrawal, and multiple among adolescents is considerably different methods and a decline in nonuse. The overall pregnancy risk index declined 38%, from the pattern in non–English-speaking Eu- with 86% of the decline attributable to improved contraceptive use. Among ado- lescents aged 15 to 17 years, 77% of the decline in pregnancy risk was attributa- and then dropped dramatically.1 Little of the Conclusions. The decline in US adolescent pregnancy rates appears to be fol- decline in Europe seems attributable to delay lowing the patterns observed in other developed countries, where improved con- in initiation of sexual intercourse, given that traceptive use has been the primary determinant of declining rates. (Am J Pub- the median age at initiation has fallen since lic Health. 2007;97:150–156. doi:10.2105/AJPH.2006.089169) 1965, indicating that more teens were havingsex.1,5 In fact, the age at which young people abstinence from sexual intercourse.13 Conse- of US high-school students in an attempt to initiate sexual activity has become increas- understand declining adolescent pregnancy ingly similar across developed countries.1,5 A abstinence until marriage (“abstinence only”) rates.16 We found significant increases in use mid-1990s analysis of 5 developed countries as its primary prevention message for teen- of contraception among 15- to 17-year-olds showed that adolescents in the United States agers.14 Federal government requirements for initiated sexual activity at an age similar to abstinence-only programs specify that these improved contraceptive use and delay in initi- that of adolescents in Sweden, France, Can- programs must have as their “exclusive pur- ation of intercourse made equal contributions ada, and Great Britain but that they used con- pose” the promotion of abstinence outside of marriage and that they must not, in any way, In an effort to update that study, we con- advocate contraceptive use or discuss contra- are the result of shifts in 2 key underlying be- the roles of increased contraceptive use and haviors: sexual activity and contraceptive use.
delayed initiation of sexual activity in ex- Between 1971 and 1988, age at sexual initia- Federal government funding for abstinence- plaining changes in pregnancy risk over the tion among US teenagers became increasingly only education in the United States has grown younger, as demonstrated by increases in the rapidly since 1998, despite a lack of scientific aged 15 to 19 years. We used data from the proportion of adolescents who had ever expe- evidence in support of these programs and rienced coitus.7–9 At the beginning of the concerns about their informational content Survey of Family Growth (NSFG), a nation- 1990s this trend reversed, and declines in and ethical acceptability.13,14 In addition, the ally representative household survey that early sexual experience have since been doc- federal government, through its foreign aid provides more complete coverage of female umented in both school-based and household adolescents (particularly older adolescents nence as a means of preventing HIV infection and those who are out of school) than high- Social conservatives in the United States school surveys. The NSFG also provides de- In a previous analysis, we examined nation- tailed information about contraceptive use, in adolescent pregnancy rates to increased ally representative data derived from samples allowing assessment of trends in dual- and 150 | Research and Practice | Peer Reviewed | Santelli et al. American Journal of Public Health | January 2007, Vol 97, No. 1 multiple-method use, which can greatly re- Data on pregnancies. We used data on 1991 using a specific contraceptive method over a to 2000 pregnancy and birth rates obtained from the National Center for Health Statistics to for women’s first year of typical use based on compare our measure of overall pregnancy risk with actual pregnancy rates.3 The pregnancy adjusted for underreporting of abortion.18 rates for 2001 were computed using the same The NSFG is a periodic (every 7 years) na- Failure rates from the 2002 NSFG were not method employed by the National Center for tional probability survey conducted among available at the time this article was written.
Health Statistics. To estimate pregnancy rates noninstitutionalized adult (15–44 years of The failure rate for nonuse of contracep- for 2002, we calculated a linear extrapolation age) residents of the United States.17 Our tion was based on widely accepted data pro- vided by Trussell.19 We estimated failure rates were aged 15 to 19 years at the time they for combined method use at most recent in- tercourse by multiplying the method-specific We initially estimated, for both 1995 and 2002 (n = 1150). Further information about failure rates calculated for the 2 methods.
2002, the percentages of female adolescents the design of the NSFG is available elsewhere who were sexually active. We then tested for (http://www.cdc.gov/nchs/nsfg.htm.).
changes in percentage over time overall and limited our failure rate calculations to the 2 by age and race/ethnicity. Next, we measured Sexual activity and contraceptive use. We re- Risk indices. We created 2 related indexes young women had used at their most recent coded the publicly available NSFG data to in- for this study: (1) the contraceptive risk sexual intercourse, as well as the number of crease the comparability of the relevant mea- index, a weighted-average contraceptive use/ sures in the 2 waves of data collection. Our analyses were based on 2 central measures: our previously labeled weighted-average con- was assigned an individual contraceptive risk recent sexual activity and contraceptive use at traceptive failure rate index16), and (2) the score on the basis of the 2 most effective con- overall pregnancy risk index. The contracep- traceptive methods she had used at her most had engaged in vaginal intercourse at any tive risk index summarizes the overall effec- recent sexual intercourse. We used this infor- point during the 3 months before the inter- tiveness of a group’s contraceptive use and mation to calculate the mean and variance of view were defined as having been recently essentially represents pregnancy risk for the the contraceptive risk index and test for sexually active. For comparison purposes, sexually active proportion of that population by summing the product of each method-spe- 2002, both overall and separately according cific failure rate and the proportion of those experienced (i.e., had ever engaged in vaginal who are sexually active using that method at In the next part of our analysis, we calcu- their most recent sexual intercourse.18,19 In lated age- and race/ethnicity-specific changes We assessed contraceptive use at most re- these calculations, nonuse of contraception over time in overall pregnancy risk index val- was considered a “method” involving a spe- ues. We computed standard errors and tests who had been sexually active in the preced- cific risk of pregnancy. Thus, here the contra- of statistical significance using the svy series ing 3 months, reducing measurement issues ceptive risk index can be represented as fol- of commands in Stata 8.2 (Stata Corp, College related to recall. Women could report use of lows: Σ(percentage of sexually active women Station, Tex) to account for the stratified sur- up to 4 contraceptive methods in combina- using method x × CFR for method x), where vey designs.20 To calculate the mean and vari- tion at their most recent sexual intercourse or x = each specific method or method combina- ance for the overall pregnancy risk index, we tion. (The CFR for each method is reported assigned sexually active teenagers a value were pregnant at the time of the interview equal to this contraceptive risk score and as- signed those not sexually active a score of having used the contraceptive method they rizes the risk of pregnancy among all adoles- zero. Implicit in this index is the fact that ado- were using when they became pregnant (most cents (including those who are not currently lescents who were not sexually active at the were using no method); these data were col- sexually active), incorporating information time of the study, even if they had previously lected in a separate section of the interview in about both the level of recent sexual activity been sexually active, did not face a current which detailed histories were obtained.
and the level of contraceptive risk among Contraceptive failure rates. In addition to the those who were sexually active at the time of sexual activity and contraceptive use mea- the study. Thus, the overall pregnancy risk nancy risk index into its component parts to as- sures, our calculations required measures of cribe the decline in pregnancy risk from 1995 method-specific contraceptive failure rates women who were sexually active multiplied to 2002 to changes in sexual activity and (CFRs). A “typical-use” CFR is the number of changes in contraceptive use. The percentage January 2007, Vol 97, No. 1 | American Journal of Public Health Santelli et al. | Peer Reviewed | Research and Practice | 151 of the decline in pregnancy rate because of the ever engaged in sexual intercourse declined increases in sexual activity among sexually decline in sexual activity was calculated as 10% (52% to 47%; P = .035; Table 1). There was a 22% decline in the 15- to 17-year-old group (P = .003), and there was no change cluding increases in the use of individual where SA represents the percentage of sexu- time points). The number of young Hispanic methods, increases in the use of multiple ally active young women and CRI represents methods, and declines in nonuse (Table 2).
the contraceptive risk index. Similarly, the tercourse declined (P = .003), but there was percentage of the decline in pregnancy rate methods included increases in the use of con- because of improved contraceptive use was Hispanic White (P = .156) or Black (P = .415) doms (36% to 53%), birth control pills (24% to 33%), injection methods (8% to 10%), and More relevant to this analysis, rates of sex- ual activity (i.e., sexual intercourse during the ceased after its removal from the US market.
preceding 3 months) did not decline signifi- nearly identical to those obtained with an al- cantly among either 15- to 19-year-olds (41% ternative approach suggested by Preston et to 38%; P = .244) or 18- and 19-year-olds.
al.21 We used a bootstrapping procedure with Among 15- to 17-year-olds, the decline in sex- 500 iterations to calculate confidence inter- ual activity (28% to 23%) was of borderline vals (CIs) for percentage changes because of statistical significance (P = .065). Hispanic 15- and withdrawal, pills and withdrawal, and in- sexual activity and percentage changes be- to 19-year-olds exhibited a decline from 46% jection and condoms. Overall, the contracep- to 35% (P = .032). Again, no significant tive risk index declined 34% (P < .001).
change was found for non-Hispanic Whites or Blacks in that age group. In general, we found 15- to 17-year-olds were even larger than smaller changes in recent sexual intercourse changes among 15- to 19-year-olds. The rate than in history of ever having sexual inter- course, as a result of small, nonsignificant whereas pill use increased from 19% to 39%.
Nonuse declined from 35% to 14%. Use of2 or more methods rose from 12% to 33%, TABLE 1—Percentages of Young Women Aged 15–19 Years Engaging in Sexual Intercourse:
National Survey of Family Growth, 1995 and 2002
the pill and condom simultaneously (22%).
The contraceptive risk index declined 46% (P < .001). Although the increase in contra-ceptive use was not as dramatic among 18- History of sexual intercourse
and 19-year-olds, the decline in the contra- ceptive risk index (27%) was still consider- able (P = .004), and the percentage in which with considerable increases in the use of indi- Recent sexual intercoursea
from 38% to 58%, and use of birth control pills increased from 29% to 40%. Use of 2 or simultaneous pill and condom use rose from 9% to 17%. The contraceptive risk index de- clined 44% (P < .001). The data for non- Table 2 should be considered with caution aDefined as withing the past 3 months.
given the small sample sizes for these groupsin both years.
152 | Research and Practice | Peer Reviewed | Santelli et al. American Journal of Public Health | January 2007, Vol 97, No. 1 TABLE 2—Percentages of Sexually Active Young Women Aged 15–19 Years Who Used Selected Contraceptive Methods at Most Recent Sexual
Intercourse and Contraceptive Failure Rates Risk Scores: National Survey of Family Growth, 1995 and 2002
Note. Typical-use first-year contraceptive failure rates are from Ranjit et al.18 unless otherwise noted.
aFrom Trussell.19b Weighted-average contraceptive use or nonuse risk score, abbreviated as contraceptive risk index.
As described in the “Methods” section, the Blacks and Hispanics, changes were of border- changes in the 2 key components: sexual ac- overall pregnancy risk index combined the line statistical significance for both groups.
tivity and contraceptive use. As Table 4 dem- impact of changes in sexual activity and con- (Note that, in each case, the decline in actual onstrates, the largest changes in behaviors traceptive use (Table 3). Overall, pregnancy birth and pregnancy rates fell within the con- fidence intervals for the change in pregnancy 15- to 17-year-olds. This finding is consistent risk. This represents one way to validate the with the largest changes in actual pregnancy among 15- to 17-year-olds (55%, from 9.7 to calculation of our overall pregnancy risk rates occurring among younger teenagers.
4.4) than among 18- and 19-year-olds (27%, from 19.6 to 14.4). The change in the overall served among 15- to 19-year-olds was attrib- utable to a decrease in the percentage of sex- Hispanic Whites was significant; however, nancy risk and also displays the overall per- ually active young women (95% CI = –18%, centages of change that could be attributed to January 2007, Vol 97, No. 1 | American Journal of Public Health Santelli et al. | Peer Reviewed | Research and Practice | 153 TABLE 3—Changes in Pregnancy Risk, by Age and Race/Ethnicity: National Survey of Family Growth,
1995 and 2002
Change, 1995–2002, % (95% confidence interval) aData for 2002 not available; change extrapolated from trend between 1995 and 2001.
TABLE 4—Summary of Changes in Sexual Activity and Risk Index Values and Overall Changes Attributable
to Sexual Activity and Contraceptive Use: National Survey of Family Growth, 1995 and 2002
Overall change attributable to sexual activity (95% CI) Overall change attributable to contraceptive use (95% CI) Note. CI = confidence interval.
CI = 74%, 128%). As noted earlier, attribu- availability and increased use of modern con- tions for non-Hispanic Blacks and Hispanics traceptives have been primarily responsible (Table 4) should be interpreted with caution for declines in adolescent pregnancy rates.1 given the limited sample sizes and large con- Our findings raise questions about current CI = 55%, 106%). (Confidence intervals for US government policies that promote absti- attributions [and the attributions themselves] nence from sexual activity as the primary strategy to prevent adolescent pregnancy.
because one of the 2 changes may have actu- ally been in the opposite direction of the Our data suggest that declining adolescent federal government’s efforts to promote overall change. For example, if sexual activity pregnancy rates in the United States between abstinence-only strategies. The limited evalu- actually increased in one group but contra- 1995 and 2002 were primarily attributable ations of abstinence-only sex education pro- ceptive use and the overall pregnancy risk to improved contraceptive use. The decline declined, sexual activity would have made a in pregnancy risk among 18- and 19-year-olds successful in delaying initiation of sexual in- “negative” contribution to the decline in preg- was entirely attributable to increased contra- tercourse.22 Although abstinence is theoreti- nancy risk, and contraceptive use would have ceptive use. Decreased sexual activity was re- cally highly effective in preventing unin- been responsible for “more than” 100% of sponsible for about one quarter (23%) of the tended pregnancies and sexually transmitted the change.) All of the change in pregnancy decline among 15- to 17-year-olds, and in- infections (STIs), in actual practice abstinence risk among 18- and 19-year-olds was the re- creased contraceptive use was responsible for intentions often fail.14,23 Abstinence pro- the remainder (77%). Improved contraceptive use included increases in the use of many other prevention behaviors. For example, a individual methods, increases in the use of longitudinal examination of the virginity mated that 7% of the change was attributa- multiple methods, and substantial declines ble to a decrease in the percentage of sexu- delay initiation of sexual intercourse; how- ally active young women (95% CI = –28%, These data suggest that the United States ever, they were less likely to use contracep- tion when they initiated sexual activity and other developed countries where increased were less likely to seek STI screenings.24 154 | Research and Practice | Peer Reviewed | Santelli et al. American Journal of Public Health | January 2007, Vol 97, No. 1 Identifying changes in the behaviors that result of differences in age groups and time questions about contraception use at most re- result in adolescent pregnancy can provide periods, inclusion of young people who are cent intercourse did assess consistency of some insight into the social forces that influ- not in school, and more complete measure- ence these behaviors. Increases in the use of changes in biological fecundity among teen- multiple methods of contraception suggest an increased motivation to avoid pregnancy and STIs, which in turn suggest declines in the so- cial acceptability of adolescent childbearing self-reported information is used, one must al- and increases in educational and employment ways consider the potential for over- and our results? Although more adolescents in the opportunities. Increasing rates of condom use under-reporting. Adolescents are generally re- United States are delaying initiation of sexual in the United States reflect continuing con- liable reporters of information on sexual intercourse, the impact of this change on health.31 However, given increasing social pregnancy risk is small and confined to youn- pressure to delay sexual initiation and avoid ger teenagers (i.e., 15- to 17-year-olds). Over- pregnancy, adolescents may be more likely all, increasing rates of contraceptive use ap- their communities may increasingly see ado- today than in the past to underreport sexual lescent pregnancy as a barrier to improve- activity or overreport contraceptive use.
declining pregnancy rates between 1995 and ments in life circumstances.26 Adolescents 2002, and this assessment appears to be con- who are also parents have become less so- adequate, sample sizes become problemati- sistent with the pattern in other developed cially acceptable.27 Delays in initiation of sex- cally small in analyses of subgroups. This was countries. Public policies and programs in ual activity are traceable to many factors, in- particularly true for the Black and Hispanic the United States and elsewhere should vigor- cluding broad public support for delaying subgroups, in which the numbers of sexually ously promote provision of accurate informa- initiation of sexual intercourse at least until tion on contraception and on sexual behavior graduation from high school.27 Ironically, the over, variance around changes in percentages and relationships, support increased availabil- trend toward later initiation of sexual inter- or around attribution was much larger than ity and accessibility of contraceptive services course and declines in adolescent pregnancy variance around estimates for a single point and supplies for adolescents, and promote the appears to have preceded recent intensive ef- in time. As such, care should be taken in in- value of responsible and protective behaviors, forts on the part of the US government to terpreting our estimates for these smaller including condom and contraceptive use and Abstinence promotion is a worthwhile goal, hensive information on the factors underlying with instability in the NSFG data for Hispanic particularly among younger teenagers; how- recent declines in US rates of adolescent adolescents. In our analyses, the decline in ever, the scientific evidence shows that, in it- pregnancy. Earlier studies involving NSFG sexual experience among Hispanic teenagers self, it is insufficient to help adolescents pre- data28,29,30 focused on the years 1988 to 1995, a period in which there were relatively larger than the changes observed in other emphasis of US domestic and global policies, small changes in rates of adolescent preg- groups. Likewise, a comparison of the 1988, which stress abstinence-only sex education to 1995, and 2002 versions of the NSFG10 re- the exclusion of accurate information on con- allow exploration of behavioral changes dur- vealed wide differences over time in sexual traception, is misguided. Similar approaches experience estimates among young Hispanic should not be adopted by other nations.
25%, respectively). These differences seem implausible and may have resulted from the The authors are with the Guttmacher Institute, New York, that both increased abstinence and increased limited sample size or other problems in- NY. John S. Santelli is also with the Heilbrunn Department contraceptive use contributed to the decline volved in sampling an ethnic group that is of Population and Family, Mailman School of PublicHealth, Columbia University, New York, NY. heterogeneous with respect to national origin Requests for reprints should be sent to John S. Santelli, olds.16 Relative to school surveys, the NSFG MD, MPH, Heilbrunn Department of Population and includes more data on older teenagers and Family Health, Mailman School of Public Health, Colum-bia University, 60 Haven Ave, B-2, New York, NY 10032 those who have left school and collects more whether contraceptives were used correctly detailed information about contraceptive use.
or in biological fecundity. Correct use of con- This article was accepted August 25, 2006. In comparison with our school-based study, this analysis of the NSFG showed a larger contribution of contraceptive use to declines most recent available failure rates (for 1995).
J. S. Santelli originated the study and assumed primaryresponsibility for the writing of the article. L. Duberstein in adolescent pregnancy rates. We believe Ranjit et al. found no changes between 1988 Lindberg was the primary data analyst and was in- that these differences in attribution are the volved in the origination of the study. L. B. Finer January 2007, Vol 97, No. 1 | American Journal of Public Health Santelli et al. | Peer Reviewed | Research and Practice | 155 provided expertise on advanced statistical methods.
15. Access to Condoms and HIV/AIDS Information: A Behind the Mask
S. Singh provided expertise on research methods and Global Health and Human Rights Concern. New York, How the World Survived SARS,
16. Santelli JS, Abma J, Ventura S, et al. Can changes the First Epidemic of the
in sexual behaviors among high school students ex- Twenty–First Century
This study was supported by the Ford Foundation plain the decline in teen pregnancy rates in the 1990s? through the Guttmacher Institute (grant 1055–0169).
J Adolesc Health. 2004;35:80–90.
By Timothy J. Brookes, in collabora- 17. Groves RM, Benson G, Mosher WD, et al. Plan and operation of cycle 6 of the National Survey of The institutional review board at Columbia University Family Growth. Vital Health Stat 1. 2005(42).
declared this study exempt from protocol approval be- 18. Ranjit N, Bankole A, Darroch JE, Singh S. Contra- ceptive failure in the first two years of use: differencesacross socioeconomic subgroups. Fam Plann Perspect. Teitler JO. Trends in youth sexual initiation and 19. Trussell J. Contraceptive efficacy. In: Hatcher RA, fertility in developed countries: 1960–1995. Ann Am Trussell J, Stewart F, et al., eds. Contraceptive Technol- Acad Political Sci. 2002;580:134–152.
ogy. New York, NY: Ardent Media Inc; 2004: Singh S, Darroch JE. Adolescent pregnancy and childbearing: levels and trends in developed countries.
20. Kohler U, Kreuter F. Data Analysis Using Stata.
Fam Plann Perspect. 2000;32:14–23.
College Station, Tex: Stata Corp; 2005.
Ventura SJ, Abma JC, Mosher WD, Henshaw S.
21. Preston SH, Heuveline P, Guillot M. Demography: Estimated pregnancy rates for the United States, Measuring and Modeling Population Processes. Oxford, 1990–2000: an update. Natl Vital Stat Rep. June 15, Martin JA, Hamilton BE, Sutton PD, Ventura SJ, 22. Kirby D, Laris B, Rolleri L. The Impact of Sex and Menacker F, Munson ML. Births: final data for 2003.
HIV Education Programs in Schools and Communities on Natl Vital Stat Rep. September 8, 2005;54(2).
Sexual Behaviors Among Young Adults. Research TrianglePark, NC: Family Health International; 2006.
Sandfort T, Hubert M, Bajos N, Bos H. Sexual be- havior and HIV risk: common patterns and differences 23. Pinkerton SD. A relative risk-based, disease-specific between European countries. In: Hubert M, Bajos N, definition of sexual abstinence failure rates. Health Educ This exciting new book tells the story of the recent outbreak of SARS—SevereAcute Respiratory Syndrome, the first Sandfort T, eds. Sexual Behavior and HIV/AIDS in Eu- rope. London, England: UCL Press; 1998:403–426.
true epidemic of the 21st century. It follows 24. Bruckner H, Bearman PS. After the promise: the Darroch JE, Singh S, Frost JJ. Differences in teen- STD consequences of adolescent virginity pledges. point of view of the virus–hunters as well as age pregnancy rates among five developed countries: J Adolesc Health. 2005;36:271–278.
the role of sexual activity and contraceptive use. Fam 25. Ku L, Sonenstein FL, Pleck JH. The dynamics of Plann Perspect. 2003;33:244–250, 281.
ISBN 0-87553-046-X • Softcover • 2004 • 262 pages
young men’s condom use during and across relation- $21.00 APHA Members • $25.00 Nonmembers Zelnik M, Kantner JF. Sexual activity, contracep- ships. Fam Plann Perspect. 1994;26:246–251.
tive use and pregnancy among metropolitan-area teen- 26. Cassell C, Santelli J, Gilbert BC, Dalmat M, Mezoff J, American Public Health
agers: 1971–1979. Fam Plann Perspect. 1980;12:230– Association
Schauer M. Mobilizing communities: an overview of 800 I Street, NW,
the Community Coalition Partnership Programs for the Forrest JD, Singh S. The sexual and reproductive Washington, DC 20001
Prevention of Teen Pregnancy. J Adolesc Health. 2005; behavior of American women, 1982–1988. Fam Plann www.apha.org
27. Albert B. With One Voice: America’s Adults and Premarital sexual experience among adolescent ORDER TODAY!
Teens Sound Off About Teen Pregnancy. Washington, women—United States, 1970–1988. MMWR Morb DC: National Campaign to Prevent Teen Pregnancy; American Public Health Association
Mortal Wkly Rep. 1991;39:929–932.
Publication Sales
10. Abma JC, Martinez GM, Mosher WE, Dawson BS.
Web: www.apha.org
28. Flanigan C. What’s Behind the Good News: The Teenagers in the United States: sexual activity, contra- E-mail: [email protected]
Decline in Teen Pregnancy Rates During the 1990s. ceptive use and childbearing, 2002. Vital Health Stat Tel: 888-320-APHA
Washington, DC: National Campaign to Prevent Teen FAX: 888-361-APHA
11. Brener N, Lowry R, Kann L, et al. Trends in sex-ual risk behaviors among high school students—United 29. Darroch JE, Singh S. Why Is Teenage Pregnancy States, 1991–2001. MMWR Morb Mortal Wkly Rep. Declining? The Roles of Abstinence, Sexual Activity and Contraceptive Use. New York, NY: Alan Guttmacher In-stitute; 1999.
12. Grunbaum JA, Kann L, Kinchen K, et al. Youthrisk behavior surveillance—United States, 2003.
30. Mohn JK, Tingle LR, Finger R. An analysis of the MMWR Morb Mortal Wkly Rep. 2004;53(SS-2):1–29.
causes of the decline in non-marital birth and preg-nancy rates for teens from 1991 to 1995. Adolesc Fam 13. Dailard C. Understanding “abstinence”: implica- tions for individuals, programs and policies. GuttmacherRep Public Policy. 2003;6:4–6.
31. Brener ND, Collins JL, Kann L, Warren CW, 14. Santelli JS, Ott MA, Lyon M, Rogers J, Summers D, Williams BI. Reliability of the Youth Risk Behavior Schleifer R. Abstinence and abstinence-only education: Survey questionnaire. Am J Epidemiol. 1995;141: a review of US policies and programs. J Adolesc Health. 156 | Research and Practice | Peer Reviewed | Santelli et al. American Journal of Public Health | January 2007, Vol 97, No. 1

Source: http://dev-www.ined.fr/fichier/t_rendezvous/459/rendezvous_fichier_en_santelli.pdf