Referenced Articles

Att I Referenced articles.pdf

Assessment of a Preventive Service Program in the Context of a Zika Virus Outbreak in Puerto Rico

Referenced Articles

OMB: 0920-1241

Document [pdf]

Download: pdf | pdf

Morbidity and Mortality Weekly Report

Zika Virus Infection Among U.S. Pregnant Travelers —
August 2015–February 2016
Dana Meaney-Delman, MD1; Susan L. Hills, MBBS2; Charnetta Williams, MD3,4; Romeo R. Galang, MD3,4; Preetha Iyengar, MD5; Andrew K.
Hennenfent, DVM6; Ingrid B. Rabe, MBChB2; Amanda Panella, MPH2; Titilope Oduyebo, MD3,7; Margaret A. Honein, PhD8; Sherif Zaki, MD, PhD9;
Nicole Lindsey, MS2; Jennifer A. Lehman2; Natalie Kwit, DVM3; Jeanne Bertolli, PhD4; Sascha Ellington, MSPH7; Irogue Igbinosa, MD10; Anna A.
Minta, MD3,11; Emily E. Petersen, MD7; Paul Mead, MD2; Sonja A. Rasmussen, MD12; Denise J. Jamieson, MD7

On February 26, 2016, this report was posted as an MMWR
Early Release on the MMWR website (http://www.cdc.gov/mmwr).
After reports of microcephaly and other adverse pregnancy
outcomes in infants of mothers infected with Zika virus during pregnancy, CDC issued a travel alert on January 15, 2016,
advising pregnant women to consider postponing travel to areas
with active transmission of Zika virus. On January 19, CDC
released interim guidelines for U.S. health care providers caring
for pregnant women with travel to an affected area (1), and
an update was released on February 5 (2). As of February 17,
CDC had received reports of nine pregnant travelers with
laboratory-confirmed Zika virus disease; 10 additional reports
of Zika virus disease among pregnant women are currently
under investigation. No Zika virus–related hospitalizations
or deaths among pregnant women were reported. Pregnancy
outcomes among the nine confirmed cases included two
early pregnancy losses, two elective terminations, and three
live births (two apparently healthy infants and one infant
with severe microcephaly); two pregnancies (approximately
18 weeks’ and 34 weeks’ gestation) are continuing without
known complications. Confirmed cases of Zika virus infection
were reported among women who had traveled to one or more
of the following nine areas with ongoing local transmission of
Zika virus: American Samoa, Brazil, El Salvador, Guatemala,
Haiti, Honduras, Mexico, Puerto Rico, and Samoa. This report
summarizes findings from the nine women with confirmed
Zika virus infection during pregnancy, including case reports
for four women with various clinical outcomes. U.S. health care
providers caring for pregnant women with possible Zika virus
exposure during pregnancy should follow CDC guidelines for
patient evaluation and management (1,2). Zika virus disease
is a nationally notifiable condition. CDC has developed a
voluntary registry to collect information about U.S. pregnant
women with confirmed Zika virus infection and their infants.
Information about the registry is in preparation and will be
available on the CDC website.
Zika virus is a mosquito-borne flavivirus that was first isolated from a rhesus monkey in Uganda in 1947 (3). For several
decades, only sporadic human disease cases were reported
from Africa and Southeast Asia. In 2007, an outbreak was
reported on Yap Island, Federated States of Micronesia (3),

and outbreaks subsequently were reported from several Pacific
Island countries (4). Local transmission of Zika virus was first
identified in the Region of the Americas (Americas) in Brazil in
May 2015 (5). Since that time, transmission of Zika virus has
occurred throughout much of the Americas; as of February 18,
a total of 32 countries and territories worldwide have active
transmission of Zika virus (http://www.cdc.gov/zika/geo/
active-countries.html). Interim guidelines for evaluation and
management of pregnant women who have traveled to areas
with ongoing local transmission of Zika virus include offering
laboratory testing after return from travel (2).
During August 1, 2015–February 10, 2016, CDC received
257 requests for Zika virus testing for pregnant women. Among
these requests, 151 (59%) included information indicating
that the woman had a clinical illness consistent with Zika virus
disease (i.e., two or more of the following signs or symptoms:
acute onset of fever, rash, conjunctivitis, or arthralgia). The
remaining requests did not document an illness compatible
with Zika virus disease, but reporting of symptom information
might have been incomplete.
Laboratory confirmation of recent Zika virus infection
includes detection of 1) Zika virus, viral RNA, or viral antigen,
or 2) Zika virus immunoglobulin M (IgM) antibodies with
Zika virus neutralizing antibody titers ≥4-fold higher than
neutralizing antibody titers against dengue or other flaviviruses
endemic to the region where exposure occurred. Among the
257 pregnant women whose specimens were tested at CDC,
249 (97%) tested negative for recent Zika virus infection and
eight (3%) had confirmed Zika virus infection. In addition to
the eight patients with laboratory testing performed at CDC,
one confirmed case was reported to CDC from a state health
department with capacity to test for Zika virus infection.
Among nine pregnant women with confirmed Zika virus
disease, no hospitalizations or deaths were reported. All nine
women reported at least one of the four most commonly
observed symptoms (fever, rash, conjunctivitis, or arthralgia),
all women reported rash, and all but one woman had at
least two symptoms. Among the six pregnant women with
Zika virus disease who reported symptoms during the first
trimester, outcomes included two early pregnancy losses,
two elective pregnancy terminations, and delivery of a live

US Department of Health and Human Services/Centers for Disease Control and Prevention

MMWR / March 4, 2016 / Vol. 65 / No. 8

211

Morbidity and Mortality Weekly Report

born infant with microcephaly; one pregnancy is continuing. Among two women with Zika virus infection who had
symptoms during the second trimester of pregnancy, one
apparently healthy infant has been born and one pregnancy
is continuing. One pregnant woman reported symptoms of
Zika virus infection in the third trimester of pregnancy, and
she delivered a healthy infant.

Selected Case Reports
Patient A. In January 2016, a pregnant woman in her 30s
reported symptoms of fever, rash, arthralgia, myalgia, and
malaise at 6–7 weeks’ gestation. She had traveled to a Zikaaffected area at approximately 5 weeks’ gestation. Serologic
testing confirmed recent Zika virus infection. She experienced
a spontaneous early pregnancy loss and underwent a dilation
and curettage at approximately 8 weeks’ gestation. Products of
conception were sent to CDC for testing, and Zika virus RNA
was detected by reverse transcription-polymerase chain reaction (RT-PCR) and immunohistochemical (IHC) staining (6).
Patient B. In January 2016, a pregnant woman in her 30s
underwent laboratory testing for Zika virus infection. She
reported a history of travel to a Zika-affected area at approximately 11–12 weeks’ gestation. One day after returning from
travel, she developed fever, eye pain, and myalgia. The next
day, she developed a rash. Serologic testing confirmed recent
Zika virus infection. At approximately 20 weeks’ gestation,
she underwent a fetal ultrasound that suggested absence of
the corpus callosum, ventriculomegaly, and brain atrophy;
subsequent fetal magnetic resonance imaging demonstrated
severe brain atrophy. Amniocentesis was performed, and Zika
virus RNA was detected by RT-PCR testing. After discussion
with her health care providers, the patient elected to terminate
her pregnancy.
Patient C. In late 2015, a woman in her 30s gave birth to
an infant at 39 weeks’ gestation. The infant’s head circumference at birth was 27 cm (<3rd percentile), indicating severe
microcephaly (http://www.cdc.gov/growthcharts/who_charts.
htm). After delivery, an epidemiologic investigation revealed
that the woman had resided in Brazil until 12 weeks’ gestation.
She reported that she had experienced fever, rash, arthralgia,
and headache at 7–8 weeks’ gestation. Evidence of Zika virus
infection in the mother was confirmed by serologic testing.
Molecular and pathologic evaluation of the placenta demonstrated Zika virus RNA by RT-PCR and IHC, respectively.
The infant exhibited hypertonia, difficulty swallowing, and
seizures, and computerized tomography scan demonstrated
multiple scattered and periventricular brain calcifications.
Funduscopic examination revealed a pale optic nerve and
mild macular chorioretinitis. Newborn hearing screening was

212

MMWR / March 4, 2016 / Vol. 65 / No. 8

normal. The infant was discharged from the hospital with a
gastrostomy feeding tube.
Patient D. A pregnant woman in her 30s traveled to a Zikaaffected area at approximately 15 weeks’ gestation. She reported
symptoms of fever, rash, arthralgia, and headache beginning at
the end of her travel (at approximately 17–18 weeks’ gestation).
Serologic testing confirmed evidence of Zika virus infection. At
approximately 40 weeks’ gestation, she delivered a full-term,
apparently healthy infant with no reported abnormalities and
a head circumference of 34.5 cm. Cranial ultrasound, newborn
hearing screen, and ophthalmologic examination of the infant
were all normal.
Discussion

On January 19, 2016, CDC released interim guidelines
recommending that pregnant women who had traveled to
areas with ongoing local transmission of Zika virus and who
had symptoms consistent with Zika virus disease be tested
for Zika virus infection (1). These guidelines were updated
and expanded on February 5 to offer Zika virus testing to
all pregnant women with Zika virus exposure, regardless of
the presence of symptoms (2). Although Zika virus testing
can be performed in some state, territorial, and local health
departments, most testing before mid-February 2016 was
performed at CDC. Based on tests performed at CDC as of
February 17, 2016, only a small number of pregnant women
who reported clinical illness consistent with Zika virus disease
had laboratory evidence of a recent Zika virus infection. The
combination of clinical signs and symptoms consistent with
suspected Zika virus disease, including fever, rash, conjunctivitis, and arthralgia, is not specific to Zika virus disease; there
are other causes of this clinical presentation (7). Among the
nine pregnant women with Zika virus infection, all reported
a clinical illness, including eight women with ≥2 signs and/or
symptoms, and one with a generalized rash. The finding of
reported clinical illness among all women who tested positive
for Zika virus might be related to the initial testing criteria for
pregnant women recommended by CDC, which required the
presence of clinical illness consistent with Zika virus disease.
Additional testing performed as of February 24, 2016 identified no confirmed cases among 162 pregnant women without
reported symptoms.
Two women with confirmed Zika virus infection experienced
spontaneous pregnancy losses in the first trimester of pregnancy. Although Zika virus RNA was detected in the specimens
from both of these cases, it is not known whether Zika virus
infection caused the pregnancy losses. First trimester pregnancy
loss is common, occurring in approximately 9%–20% of all
clinically recognized pregnancies (8), with higher rates in
older women. Pregnancy loss has been observed in association

US Department of Health and Human Services/Centers for Disease Control and Prevention

Morbidity and Mortality Weekly Report

Summary
What is already known about this topic?
Because of the risk for Zika virus infection and its possible
association with adverse pregnancy outcomes, CDC issued a
travel alert on January 15, 2016, advising pregnant women to
consider postponing travel to areas with ongoing local
transmission of Zika virus. CDC also released guidelines for Zika
virus testing for pregnant women with a history of travel while
pregnant to areas with ongoing Zika virus transmission.
What is added by this report?
This report provides preliminary information on testing for Zika
virus infection of U.S. pregnant women who had traveled to areas
with Zika virus transmission. As of February 17, 2016, nine U.S.
pregnant travelers with Zika virus infection had been identified.
No Zika virus–related hospitalizations or deaths were reported
among pregnant women. Pregnancy outcomes included two
early pregnancy losses, two elective terminations, and three live
births (two apparently healthy infants and one infant with severe
microcephaly); two pregnancies (18 weeks’ and 34 weeks’
gestation) are continuing without known complications.
What are the implications for public health practice?
In this small case series, Zika virus infection during pregnancy
was associated with a range of outcomes, including early
pregnancy losses, congenital microcephaly, and apparently
healthy infants. Additional information will be available in the
future from a newly established CDC registry for U.S. pregnant
women with confirmed Zika virus infection and their infants.

with Zika virus infection (6) and after infections with other
flaviviruses (e.g., dengue, West Nile, Japanese encephalitis)
(9–11); however, a causal relationship has not been established.
Additional histopathologic evaluation and RT-PCR testing of
tissues from pregnancy losses might provide additional insight
into maternal-fetal transmission of Zika virus and the link
between maternal-fetal transmission and pregnancy losses.
Seven pregnant women with confirmed Zika virus infection
reported fever during pregnancy. Fever has been determined
to increase the risk for adverse pregnancy outcomes, including
neural tube defects (12). It is not known whether fever might
have affected pregnancy outcomes among these pregnant
women with Zika virus infection. Because of the potential
risks for poor outcomes associated with fever during pregnancy, acetaminophen should be used to treat fever during
pregnancy (12).
Approximately half a million pregnant women are estimated
to travel to the United States annually from the 32 (as of
February 18, 2016) Zika-affected countries and U.S. territories
with active transmission of Zika virus (personal communication, Bradley Nelson, February 23, 2016). These numbers
might decrease if pregnant women follow CDC recommendations (1) and postpone travel to areas with ongoing local

Zika virus transmission. Pregnant women and their partners
should also be aware of the risk for Zika virus infection through
unprotected sex with an infected male partner, and carefully
follow CDC interim guidelines for preventing sexual transmission of Zika virus infection (13). Health care providers should
notify their state, local, or territorial health department about
women with possible exposure to Zika virus during pregnancy
for assistance in arranging testing and interpreting results.
CDC has developed a registry to collect information on U.S.
pregnant women with confirmed Zika virus infection and their
infants. Information gathered from public health officials or
health care providers will include clinical information about
the pregnancy and the infant at birth and through the first
year of life. This voluntary registry has been determined to be a
nonresearch public health surveillance activity, and as such, it is
not subject to institutional review board requirements. Health
care providers are encouraged to discuss participation in the
U.S. registry* with pregnant women with Zika virus infection.
* For inquiries about the U.S. Pregnancy Registry, please contact the
corresponding author.

Acknowledgments
Noreen A. Hynes, MD, Johns Hopkins University Schools of
Medicine and Public Health; Roberta L. DeBiasi MD, Children’s
National Medical Center, George Washington University School of
Medicine; Richard Kennedy, MD, One Medical Group.
1Office of the Director, National Center for Emerging and Zoonotic Infectious
Diseases, CDC; 2Division of Vector-Borne Diseases, National Center for
Emerging and Zoonotic Infectious Diseases, CDC; 3Epidemic Intelligence
Service, CDC; 4Division of HIV/AIDS Prevention, National Center for HIV/
AIDS, Viral Hepatitis, STD, and Tuberculosis Prevention, CDC; 5District of
Columbia Department of Health; 6CDC/CSTE Applied Epidemiology
Fellowship, District of Columbia Department of Health; 7Division of
Reproductive Health, National Center for Chronic Disease Prevention and
Health Promotion, CDC; 8Division of Congenital and Developmental
Disorders, National Center on Birth Defects and Developmental Disabilities,
CDC; 9Division of High Consequence Pathogens, National Center for
Emerging and Zoonotic Infectious Diseases, CDC; 10Division of Scientific
Education and Development, CDC; 11Division of Parasitic Diseases and
Malaria, Center for Global Health, CDC; 12Division of Public Health
Information Dissemination, Center for Surveillance, Epidemiology, and
Laboratory Services, CDC.
Corresponding author: Dana Meaney-Delman, [email protected],
770-488-7100.

References
1. Petersen EE, Staples JE, Meaney-Delman D, et al. Interim guidelines
for pregnant women during a Zika virus outbreak—United States, 2016.
MMWR Morb Mortal Wkly Rep 2016;65:30–3. http://dx.doi.
org/10.15585/mmwr.mm6502e1
2. Oduyebo T, Petersen EE, Rasmussen SA, et al. Update: interim guidelines
for health care providers caring for pregnant women and women of
reproductive age with possible Zika virus exposure—United States, 2016.
MMWR Morb Mortal Wkly Rep 2016;65:122–7. http://dx.doi.
org/10.15585/mmwr.mm6505e2

US Department of Health and Human Services/Centers for Disease Control and Prevention

MMWR / March 4, 2016 / Vol. 65 / No. 8

213

Morbidity and Mortality Weekly Report

3. Duffy MR, Chen T-H, Hancock WT, et al. Zika virus outbreak on Yap
Island, Federated States of Micronesia. N Engl J Med 2009;360:2536–43.
http://dx.doi.org/10.1056/NEJMoa0805715
4. Musso D, Nilles EJ, Cao-Lormeau VM. Rapid spread of emerging Zika
virus in the Pacific area. Clin Microbiol Infect 2014;20:O595–6. http://
dx.doi.org/10.1111/1469-0691.12707
5. Hennessey M, Fischer M, Staples JE. Zika virus spreads to new areas—
Region of the Americas, May 2015–January 2016. MMWR Morb Mortal
Wkly Rep 2016;65:55–8. http://dx.doi.org/10.15585/mmwr.mm6503e1
6. Martines RB, Bhatnagar J, Keating MK, et al. Notes from the field:
evidence of Zika virus infection in brain and placental tissues from two
congenitally infected newborns and two fetal losses—Brazil, 2015.
MMWR Morb Mortal Wkly Rep 2016;65:159–60. http://dx.doi.
org/10.15585/mmwr.mm6506e1
7. Roth A, Mercier A, Lepers C, et al. Concurrent outbreaks of dengue,
chikungunya and Zika virus infections—an unprecedented epidemic wave of
mosquito-borne viruses in the Pacific 2012–2014. Euro Surveill 2014;19:20929.
http://dx.doi.org/10.2807/1560-7917.ES2014.19.41.20929
8. Wilcox AJ, Weinberg CR, O’Connor JF, et al. Incidence of early loss of
pregnancy. N Engl J Med 1988;319:189–94. http://dx.doi.org/10.1056/
NEJM198807283190401

214

MMWR / March 4, 2016 / Vol. 65 / No. 8

9. Chaturvedi UC, Mathur A, Chandra A, Das SK, Tandon HO, Singh
UK. Transplacental infection with Japanese encephalitis virus. J Infect
Dis 1980;141:712–5. http://dx.doi.org/10.1093/infdis/141.6.712
10. O’Leary DR, Kuhn S, Kniss KL, et al. Birth outcomes following West Nile
virus infection of pregnant women in the United States: 2003–2004. Pediatrics
2006;117:e537–45. http://dx.doi.org/10.1542/peds.2005-2024
11. Tsai TF. Congenital arboviral infections: something new, something old.
Pediatrics 2006;117:936–9. http://dx.doi.org/10.1542/peds.2005-2729
12. Rasmussen SA, Jamieson DJ, Macfarlane K, Cragan JD, Williams J,
Henderson Z; Pandemic Influenza and Pregnancy Working Group.
Pandemic influenza and pregnant women: summary of a meeting of
experts. Am J Public Health 2009;99(Suppl 2):S248–54. http://dx.doi.
org/10.2105/AJPH.2008.152900
13. Oster AM, Brooks JT, Stryker JE, et al. Interim guidelines for prevention
of sexual transmission of Zika virus—United States, 2016. MMWR
Morb Mortal Wkly Rep 2016;65:120–1. http://dx.doi.org/10.15585/
mmwr.mm6505e1

US Department of Health and Human Services/Centers for Disease Control and Prevention

Reproduced with permission of the copyright owner. Further reproduction prohibited without
permission.

The

n e w e ng l a n d j o u r na l

m e dic i n e

Spe ci a l R e p or t

Zika Virus and Birth Defects — Reviewing the Evidence
for Causality
Sonja A. Rasmussen, M.D., Denise J. Jamieson, M.D., M.P.H.,
Margaret A. Honein, Ph.D., M.P.H., and Lyle R. Petersen, M.D., M.P.H.

Summary
The Zika virus has spread rapidly in the Americas
since its first identification in Brazil in early 2015.
Prenatal Zika virus infection has been linked to
adverse pregnancy and birth outcomes, most notably microcephaly and other serious brain anomalies. To determine whether Zika virus infection
during pregnancy causes these adverse outcomes, we evaluated available data using criteria
that have been proposed for the assessment of
potential teratogens. On the basis of this review,
we conclude that a causal relationship exists between prenatal Zika virus infection and microcephaly and other serious brain anomalies. Evidence that was used to support this causal
relationship included Zika virus infection at times
during prenatal development that were consistent with the defects observed; a specific, rare
phenotype involving microcephaly and associated
brain anomalies in fetuses or infants with presumed or confirmed congenital Zika virus infection; and data that strongly support biologic
plausibility, including the identification of Zika
virus in the brain tissue of affected fetuses and
infants. Given the recognition of this causal relationship, we need to intensify our efforts toward
the prevention of adverse outcomes caused by
congenital Zika virus infection. However, many
questions that are critical to our prevention efforts remain, including the spectrum of defects
caused by prenatal Zika virus infection, the degree of relative and absolute risks of adverse outcomes among fetuses whose mothers were infected at different times during pregnancy, and
factors that might affect a woman’s risk of adverse pregnancy or birth outcomes. Addressing
these questions will improve our ability to reduce
the burden of the effects of Zika virus infection
during pregnancy.

P otential Rel ationship be t ween
Zik a Virus Infec tion and Bir th
Defec t s
Since the identification of the Zika virus in Brazil in early 2015, the virus has spread rapidly
throughout the Americas (www.cdc.gov/zika/
geo/active-countries.html). An increase in the
number of infants with microcephaly in Brazil
was first noted in September 2015, after the
recognition of Zika virus transmission in the
country earlier in the year1; this was followed by
the recognition of a similar increase in French
Polynesia after an outbreak there in 2013 and
2014.2 Despite accumulating evidence that supports the link between Zika virus infection and
microcephaly, most experts have taken care not
to state that Zika virus infection is causally related to these adverse outcomes.3 This cautious
approach toward ascribing Zika virus as a cause
of birth defects is not surprising, given that the
last time an infectious pathogen (rubella virus)
caused an epidemic of congenital defects was
more than 50 years ago, no flavivirus has ever
been shown definitively to cause birth defects in
humans,4 and no reports of adverse pregnancy
or birth outcomes were noted during previous
outbreaks of Zika virus disease in the Pacific
Islands.5,6
On the basis of the available evidence, the
public health response to the outbreak of Zika
virus disease has moved forward, with the distribution of health messages about the importance of mosquito-bite prevention, recommendations by public health authorities in some of the
most severely affected countries to delay pregnancy, and advisories that pregnant women avoid
travel to areas with active Zika virus transmission.7
However, communications regarding Zika virus
have been challenging: a recent survey showed

n engl j med nejm.org

The New England Journal of Medicine
Downloaded from nejm.org at Stephen B. Thacker CDC Library on April 25, 2016. For personal use only. No other uses without permission.
Copyright © 2016 Massachusetts Medical Society. All rights reserved.

The

n e w e ng l a n d j o u r na l

low levels of knowledge and concern about Zika
virus in the United States.8 The recognition of
Zika virus as a cause of microcephaly and other
serious brain anomalies would allow for more
direct communication, which might lead to improved understanding of and adherence to public
health recommendations. Therefore, a review of
the evidence linking Zika virus infection and adverse pregnancy and birth outcomes is needed.
As is typically the case in epidemiology and
medicine, no “smoking gun” (a single definitive
piece of evidence that confirms Zika virus as a
cause of congenital defects) should have been anticipated. Instead, the determination of a causal
relationship would be expected to emerge from
various lines of evidence, each of which suggests,
but does not on its own prove, that prenatal Zika
virus infection can cause adverse outcomes. Two
approaches have been used to identify potential
teratogens (exposures to a mother during pregnancy that have a harmful effect on her embryo
or fetus)9: first, the identification of a combination
of a rare exposure and a rare defect (sometimes
referred to as the astute clinician approach),10 and
second, the use of epidemiologic data to confirm
an association. Many teratogens were first identified by means of the rare exposure–rare defect
approach, including rubella virus, which was identified after an ophthalmologist noted a characteristic form of cataracts in infants whose mothers
had rubella during pregnancy,11 and heavy alcohol
use, which was identified as a teratogen after the
recognition of a characteristic pattern of malformations that became known as the fetal alcohol
syndrome.12 In contrast, some teratogens have
been identified on the basis of epidemiologic
studies (e.g., valproic acid was identified as a
teratogen after a case–control study showed an
odds ratio of 20 for the association of spina bifida
with use of this drug during the first trimester of
pregnancy).13

Shepard’s Criteria
In 1994, Thomas Shepard, a pioneer in the field
of teratology, proposed a set of seven criteria for
“proof” of human teratogenicity (Table 1) that
incorporated both approaches.9 These criteria
were an amalgamation of criteria developed by
other teratologists and guided by methods that
were used to identify previous teratogens. These
criteria have been used to guide discussions
2

m e dic i n e

about causation in teratology-related litigation30
and to assess other potential teratogens.10 We used
Shepard’s criteria9 as a framework to evaluate
whether the currently available evidence supports
the hypothesis that prenatal Zika virus infection
is a cause of microcephaly and other brain anomalies (Table 1).
According to these criteria, causality is established when either criteria 1, 3, and 4 (rare exposure–rare defect approach) or criteria 1, 2, and
3 (epidemiologic approach) are fulfilled. The first
criterion states that a proven exposure to an agent
must occur at a critical time during prenatal development. The severe microcephaly and other
brain anomalies that have been observed in many
infants are consistent with an infection occurring
in the first or early second trimester of pregnancy. Several case reports and studies have shown
that women who had fetuses or infants with
congenital brain anomalies that were believed,
on the basis of the mother’s symptoms or laboratory confirmation, to be due to Zika virus infection were infected in the first or early second
trimester of pregnancy, as determined either
according to the timing of the symptoms or according to the timing of travel to an area where
Zika virus is endemic.14-20 An analysis of the timing of laboratory-confirmed Zika virus transmission in certain states in Brazil and of the increase in the cases of microcephaly identified
the first trimester as the critical time period for
infection.1 Zika virus infections that occur later
in pregnancy have been associated with poor intrauterine growth, fetal death, or in some pregnancies, defects on prenatal imaging that have
not yet been confirmed postnatally because the
pregnancies are ongoing.14 We conclude that
Shepard’s first criterion has been met.
Shepard’s second criterion requires that two
epidemiologic studies of high quality support the
association. Although ecologic data do not necessarily qualify as an epidemiologic study, data from
Brazil regarding the temporal and geographic
association between Zika virus infection and the
later appearance of infants with congenital microcephaly are compelling.1,31,32 Two epidemiologic studies also provide support.2,14 In a study
conducted during the outbreak in Brazil, 88 pregnant women who had had an onset of rash in the
previous 5 days were tested for Zika virus RNA.
Among the 72 women who had positive tests,
42 underwent prenatal ultrasonography, and fe-

n engl j med nejm.org

Special Report

Table 1. Shepard’s Criteria for Proof of Teratogenicity in Humans as Applied to the Relationship between Zika Virus Infection and
Microcephaly and Other Brain Anomalies.*
Criterion
No.

Criterion

Evidence

Criterion Met?

Proven exposure to the agent at one or On the basis of case reports, case series, and epidemiologic studies of
more critical times during prenatal
microcephaly that are associated with laboratory-confirmed or predevelopment
sumed Zika virus infection, the timing of Zika virus infection associated with severe microcephaly and intracranial calcifications appears
to be in the late first or early second trimester.14-20

Yes

Consistent findings by ≥2 high-quality On the basis of data from Brazil, the temporal and geographic associaepidemiologic studies, with contion between Zika virus illness and cases of microcephaly is strong.1
trol of confounding factors, suffi- Two epidemiologic studies have been published. In a study in Brazil14
that used a prospective cohort design, 29% of women with Zika virus
cient numbers, exclusion of posiinfection at any time during pregnancy had abnormalities on prenatal
tive and negative bias factors, proultrasonography, some of which have not been confirmed postnatalspective studies if possible, and
ly, In a study in French Polynesia,2 retrospective identification of eight
relative risk ≥6
cases of microcephaly and the use of serologic and statistical data
and mathematical modeling suggested that 1% of fetuses and infants
born to women with Zika virus infection during the first trimester had
microcephaly; the risk ratio in this analysis was approximately 50, as
compared with the baseline prevalence of microcephaly.
No other epidemiologic studies have examined this association to date.

Partially

Careful delineation of clinical cases; a
specific defect or syndrome, if
present, is very helpful

The phenotype has been well characterized in fetuses and infants with
presumed congenital Zika virus infection, including microcephaly and
other serious brain anomalies, redundant scalp skin, eye findings, arthrogryposis, and clubfoot.15,20-23
The phenotype in some infants appears to be consistent with the fetal
brain disruption sequence,20,22 which has been observed after infection with other viral teratogens.24

Yes

Rare environmental exposure that is
associated with rare defect

Reports of fetuses and infants with microcephaly who are born to women
with brief periods of travel to countries with active Zika virus transmission are consistent with Zika virus being a rare exposure.16,18,19
The defect, congenital microcephaly, is rare, with a birth prevalence of
approximately 6 cases per 10,000 liveborn infants, according to data
from birth-defects surveillance systems in the United States.25

Yes

Teratogenicity in experimental animals No results of an animal model with Zika virus infection during pregnancy
important but not essential
and fetal effects have yet been published.

Association should make biologic
sense

Findings are similar to those seen after prenatal infection with some other viral teratogens (e.g., cytomegalovirus, rubella virus).26
Animal models have shown that Zika virus is neurotropic,27,28 which supports biologic plausibility.
Evidence that Zika virus infects neural progenitor cells and produces cell
death and abnormal growth,29 along with evidence of Zika virus in
brains of fetuses and infants with microcephaly, on the basis of immunohistochemical staining and identification of Zika virus RNA and
live virus,16,17,19 provides strong biologic plausibility.

Yes

Proof in an experimental system that This criterion applies to a medication or chemical exposure, not to infecthe agent acts in an unaltered state
tious agents.

* The criteria listed here were proposed by Shepard.9 Criteria 1, 2, and 3 or criteria 1, 3, and 4 are considered to be essential, whereas criteria
5, 6, and 7 are helpful but not essential. Partial evidence is insufficient to meet a criterion. NA denotes not applicable.

tal abnormalities were observed in 12 (29%); none
of the 16 women with negative tests had fetal
abnormalities. The abnormalities that were observed on ultrasonography varied widely, and some
findings lacked postnatal confirmation because
the pregnancies were ongoing.14
A retrospective analysis after the 2013–2014

outbreak of Zika virus disease in French Polynesia identified eight cases of microcephaly; the
authors used serologic and statistical data and
mathematical modeling to estimate that 1% of the
fetuses and neonates who were born to mothers
who had been infected with Zika virus in the
first trimester had microcephaly2 — a prevalence

n engl j med nejm.org

The

n e w e ng l a n d j o u r na l

that was approximately 50 times as high as the
estimated baseline prevalence. However, this estimate was based on small numbers, confidence
intervals were wide, and the risk of other adverse
outcomes (e.g., other brain anomalies) was not
assessed.2 Although these studies provide important evidence in support of a causal relationship between Zika virus and microcephaly and
other brain anomalies, both have limitations as
noted by their authors, such as a lack of control
for confounding factors and relatively small numbers of cases, and therefore they do not meet the
stringent criteria set by Shepard. Thus, we conclude that Shepard’s second criterion has not yet
been satisfied.
The third criterion, careful delineation of
clinical cases with the finding of a specific defect or syndrome, appears to be met. Previous
teratogens have caused specific birth defects or
syndromes rather than a broad range of birth
defects.33 Many fetuses and infants with presumed
congenital Zika virus infection have had a typical
pattern, including severe microcephaly, intracranial calcifications, and other brain anomalies,
sometimes accompanied by eye findings, redundant scalp skin, arthrogryposis, and clubfoot15,20-23;
such findings have led authors to use the term
“congenital Zika syndrome.”22,34,35 On the basis
of clinical details from a limited number of cases,
some infants with presumed congenital Zika virus infection have had features that were consistent with fetal brain disruption sequence,24 a phenotype involving the brain that is characterized by
severe microcephaly, overlapping cranial sutures,
prominent occipital bone, redundant scalp skin,
and considerable neurologic impairment.20,22 For
example, 11 of 35 infants (31%) with microcephaly whose cases were reported to a Brazil Ministry of Health registry had excessive and redundant
scalp skin,20 a finding that is not typically seen in
other forms of microcephaly.36 These findings suggest an interruption of cerebral growth, but not in
that of the scalp skin, after an injury (e.g., viral
infection, hyperthermia, or vascular disruption)
that occurred after the initial formation of brain
structures, followed by partial collapse of the
skull. The fetal brain disruption sequence is rare;
only 20 cases were identified in a literature review
in 2001.24
Shepard’s fourth criterion refers to the association between a rare exposure and a rare defect; we conclude that this criterion also has
4

m e dic i n e

been met. The concept behind this criterion is
that a rare defect occurring after a rare exposure
during pregnancy implies causation because of
the unlikelihood of the two rare events occurring together.10 Microcephaly is a rare defect that
is estimated to occur in 6 infants per 10,000 liveborn infants in the United States.25 Zika virus
would not be a rare exposure among women living
in Brazil during the Zika virus outbreak. However,
reports of adverse birth outcomes among travelers
who spent only a limited time period in an area
where there is active Zika virus transmission are
consistent with Zika virus being a rare exposure.16,18,19
A recent report is illustrative: a pregnant woman traveled for 7 days to Mexico, Guatemala, and
Belize during her 11th week of gestation and
had a positive test for Zika virus immunoglobulin M (IgM) antibodies 4 weeks later. On fetal
ultrasonography and magnetic resonance imaging performed at 19 to 20 weeks of gestation, severe brain anomalies were diagnosed in the fetus,
and the pregnancy was terminated at 21 weeks of
gestation. Microcephaly was not present at the
time of pregnancy termination, but the head circumference had decreased from the 47th percentile at 16 weeks of gestation to the 24th percentile at 20 weeks of gestation (a finding that is
consistent with the timing of diminishing head
sizes in previous cases),14 which suggests that
microcephaly would have developed in the fetus
had the pregnancy continued.16 In this woman,
Zika virus would be considered a rare exposure,
and her fetus had a rare outcome.
The last three criteria are helpful if they are
present, but they are not considered to be essential. The fifth criterion, the need for an animal
model that shows teratogenicity, has not been
met. Although animal models have shown that
Zika virus is neurotropic,27,28 no studies that tested for teratogenicity in an animal model have
been published, although studies are under way.
The sixth criterion, that the association should
make biologic sense, is clearly met here. Other
viral infections have had similar effects (microcephaly and eye problems).24,26 In addition, pathologic evidence supports this association: Zika virus RNA has been seen in damaged mononuclear
cells (presumably glial cells and neurons) in the
brains of newborns with microcephaly,17 and the
virus appears to be neurotropic.17,19 Live Zika virus has been cultured from the brain of a fetus

n engl j med nejm.org

Special Report

Table 2. Bradford Hill Criteria for Evidence of Causation as Applied to the Relationship between Zika Virus Infection
and Microcephaly and Other Brain Anomalies*
Criterion

Evidence

Criterion Met?

Strength of association

A recent epidemiologic study from French Polynesia suggests a strong association between prenatal Zika virus infection and microcephaly (estimated risk ratio, approximately 50).2
The substantial increase in the number of cases of microcephaly and other
brain anomalies that have been associated with the Zika virus outbreak in
Brazil suggests a strong association.1,2

Yes

Consistency

Two epidemiologic studies, one from Brazil and one from French Poly
nesia,2,14 support the association between prenatal Zika virus infection
and microcephaly and other serious brain anomalies.
The observed increase in the number of cases of microcephaly after outbreaks of Zika virus infection in Brazil and French Polynesia, as well as
preliminary reports of cases in Colombia, support consistency.1,2,42
Case reports of Zika virus infection in fetuses or infants with microcephaly or
other brain anomalies who were born to mothers who traveled to areas of
active Zika virus transmission support consistency.16,18,19

Yes

Specificity

Other causes of microcephaly exist; however, on the basis of clinical descriptions that are available for a small number of infants with presumed congenital Zika virus infection,20 the clinical phenotype linked to the Zika virus appears to be an unusual form of microcephaly that is consistent with
the fetal brain disruption sequence.

Yes

Temporality

Zika virus infection in mothers during pregnancy precedes the finding of microcephaly or other brain anomalies in fetuses or infants.14-20
Zika virus outbreaks in Brazil and French Polynesia preceded the increase in
the number of cases of microcephaly.1,2

Yes

Biologic gradient

Infection is a phenomenon that is either present or absent; there is no dose–
response relationship.
No data are available regarding whether women with an increased viral load
have a higher risk of adverse pregnancy or birth outcomes.

Plausibility

Findings are similar to those seen after prenatal infection with some other viral teratogens (e.g., cytomegalovirus and rubella virus).26
Evidence that Zika virus infects neural progenitor cells and produces cell
death and abnormal growth,29 along with evidence of Zika virus in brains
of fetuses and infants with microcephaly, on the basis of on immunohistochemical staining and identification of Zika virus RNA and live virus,16,17,19 provides strong biologic plausibility.

Yes

Coherence

No results in an animal model of effects of Zika virus on pregnancy have yet
been published, but animal models have shown that Zika virus is neurotropic,27,28 a finding that is consistent with prenatal Zika virus infection
causing microcephaly and other brain anomalies.
Zika virus infects neural progenitor cells and produces cell death and abnormal growth,29 a finding that is consistent with a causal relationship between Zika virus infection and microcephaly.

Yes

Experiment

No experimental animal model of Zika virus teratogenicity is available.

Analogy

No other flavivirus has been shown to definitively cause birth defects in humans,4 but flaviviruses, Wesselsbron and Japanese encephalitis viruses,
have been shown to cause stillbirth and brain anomalies in animals.43
Findings are similar to those seen after prenatal infection with other viral teratogens (e.g., cytomegalovirus, rubella virus).26

Yes

* The criteria listed here were proposed by Hill.40 We have updated a recent analysis by Frank et al.41

with severe brain anomalies after maternal infection at 11 weeks of gestation.16 Furthermore, Zika
virus efficiently infects neural progenitor cells
and produces cell death and abnormal growth,
thus providing a possible mechanism for micro-

cephaly.29 The seventh criterion, proof in an experimental system that the agent acts in an unaltered state, is aimed at medications or chemical
exposures and does not apply to infectious agents.
Thus, given Shepard’s criteria as a framework,

n engl j med nejm.org

The

n e w e ng l a n d j o u r na l

m e dic i n e

criteria 1, 3, and 4 have been satisfied — evidence preliminary reports that are being investigated
that is considered sufficient to identify an agent in Colombia.1,2,42
as a teratogen.
Moving from a hypothesis that Zika virus is
linked to certain adverse outcomes to a statement that Zika virus is a cause of certain adverse
Other Criteria
outcomes allows for direct communications reOther criteria can also be used to assess this garding risk, both in clinical care settings and
relationship. Koch’s postulates, developed in the in public health guidance, and an intensified
late 19th century, are often cited as necessary to focus on prevention efforts, such as the impleshow causation in infectious disease; however, mentation of vector control, the identification of
many authors have noted the need for Koch’s improved diagnostic methods, and the developpostulates to be updated to accommodate mod- ment of a Zika virus vaccine.44 In addition, after
ern technologies.37-39 The Bradford Hill criteria40 recognizing a causal relationship between Zika
provide another framework to assess causation; virus infection and adverse pregnancy and birth
Frank et al. recently used these criteria to assess outcomes, we can focus research efforts on other
the relationship between prenatal Zika virus in- critical issues: First, understanding the full specfection and microcephaly and concluded that trum of defects caused by congenital Zika virus
additional information was needed to assume infection; if Zika virus is similar to other teratothat the relationship was causal.41 However, sev- gens, an expansion of the phenotype would be
eral key pieces of evidence have become available expected (e.g., with the congenital rubella synsince they performed their analysis, including drome, the phenotype was expanded from catatwo epidemiologic studies,2,14 a study of the ef- racts to include other findings such as hearing
fects of Zika virus on neural progenitor cells,29 loss, congenital heart defects, and microcephaand a case report of a fetus with brain anomalies ly).11 Second, quantifying the relative and absoand decreasing head size from whose brain live lute risks among infants who are born to women
Zika virus was isolated.16 On the basis of our up- who were infected at different times during pregdate of their analysis, which incorporates newly nancy. Third, identifying factors that modify the
available evidence (Table 2), nearly all the rele- risk of an adverse pregnancy or birth outcome
vant criteria have been met, with the exception (e.g., coinfection with another virus, preexisting
of the presence of experimental evidence. How- immune response to another flavivirus, genetic
ever, Hill emphasizes that meeting all nine crite- background of the mother or fetus, and severity
ria is not necessary40; instead, the criteria should of infection). Addressing these issues will imserve as a framework to assess when the most prove our efforts to minimize the burden of the
likely interpretation of a relationship is causation. effects of Zika virus infection during pregnancy.

A sse ssment of Criteria
Thus, on the basis of a review of the available
evidence, using both criteria that are specific for
the evaluation of potential teratogens9 and the
Bradford Hill criteria40 as frameworks, we suggest that sufficient evidence has accumulated to
infer a causal relationship between prenatal Zika
virus infection and microcephaly and other severe brain anomalies. Also supportive of a
causal relationship is the absence of an alternative explanation; despite the extensive consideration of possible causes, researchers have been
unable to identify alternative hypotheses that
could explain the increase in cases of microcephaly that were observed first in Brazil and then
retrospectively in French Polynesia, and now in
6

Disclosure forms provided by the authors are available with
the full text of this article at NEJM.org.
From the Division of Public Health Information Dissemination,
Center for Surveillance, Epidemiology, and Laboratory Services
(S.A.R.), Division of Reproductive Health, National Center for
Chronic Disease Prevention and Health Promotion (D.J.J.), and
Division of Congenital and Developmental Disorders, National
Center on Birth Defects and Developmental Disabilities
(M.A.H.), Centers for Disease Control and Prevention, Atlanta;
and the Division of Vector-Borne Diseases, National Center for
Emerging and Zoonotic Infectious Diseases, Centers for Disease Control and Prevention, Fort Collins, CO (L.R.P.). Address
reprint requests to Dr. Rasmussen at the Centers for Disease
Control and Prevention, 1600 Clifton Rd., MS E-33, Atlanta, GA
30329, or at skr9@cdc.gov.
This article was published on April 13, 2016, at NEJM.org.
1. Kleber de Oliveira W, Cortez-Escalante J, De Oliveira WT, et
al. Increase in reported prevalence of microcephaly in infants
born to women living in areas with confirmed Zika virus transmission during the first trimester of pregnancy — Brazil, 2015.
MMWR Morb Mortal Wkly Rep 2016;65:242-7.

n engl j med nejm.org

Special Report

2. Cauchemez S, Besnard M, Bompard P, et al. Association be-

tween Zika virus and microcephaly in French Polynesia, 2013-15:
a retrospective study. Lancet 2016 March 15 (Epub ahead of print).
3. Petersen LR, Jamieson DJ, Powers AM, Honein MA. Zika virus. N Engl J Med 2016;374:1552-63.
4. O’Leary DR, Kuhn S, Kniss KL, et al. Birth outcomes following West Nile Virus infection of pregnant women in the United
States: 2003-2004. Pediatrics 2006;117(3):e537-45.
5. Duffy MR, Chen T-H, Hancock WT, et al. Zika virus outbreak
on Yap Island, Federated States of Micronesia. N Engl J Med
2009;360:2536-43.
6. Ioos S, Mallet HP, Leparc Goffart I, Gauthier V, Cardoso T,
Herida M. Current Zika virus epidemiology and recent epidemics. Med Mal Infect 2014;44:302-7.
7. Chang C, Ortiz K, Ansari A, Gershwin ME. The Zika outbreak of the 21st century. J Autoimmun 2016;68:1-13.
8. Associated Press-NORC Center for Public Affairs Research.
The Zika virus: Americans’ awareness and opinions of the U.S.
response. April 2016 (http://www.apnorc.org/PDFs/Zika/
2016-04%20Zika%20Virus%20Issue%20Brief%20DTPB_v1r5.pdf).
9. Shepard TH. “Proof” of human teratogenicity. Teratology
1994;50:97-8.
10. Carey JC, Martinez L, Balken E, Leen-Mitchell M, Robertson
J. Determination of human teratogenicity by the astute clinician
method: review of illustrative agents and a proposal of guidelines. Birth Defects Res A Clin Mol Teratol 2009;85:63-8.
11. Webster WS. Teratogen update: congenital rubella. Teratology 1998;58:13-23.
12. Jones KL, Smith DW. Recognition of the fetal alcohol syndrome in early infancy. Lancet 1973;302:999-1001.
13. Lammer EJ, Sever LE, Oakley GP Jr. Teratogen update: valproic acid. Teratology 1987;35:465-73.
14. Brasil P, Pereira JP Jr, Raja Gabaglia C, et al. Zika virus infection in pregnant women in Rio de Janeiro — preliminary report.
N Engl J Med. DOI:10.1056/NEJMoa1602412.
15. Calvet G, Aguiar RS, Melo AS, et al. Detection and sequencing of Zika virus from amniotic fluid of fetuses with microcephaly in Brazil: a case study. Lancet Infect Dis 2016 February
17 (Epub ahead of print).
16. Driggers RW, Ho CY, Korhonen EM, et al. Zika virus infection with prolonged maternal viremia and fetal brain abnormalities. N Engl J Med. DOI:10.1056/NEJMoa1601824.
17. Martines RB, Bhatnagar J, Keating MK, et al. Notes from the
field: evidence of Zika virus infection in brain and placental tissues
from two congenitally infected newborns and two fetal losses —
Brazil, 2015. MMWR Morb Mortal Wkly Rep 2016;65:159-60.
18. Meaney-Delman D, Hills SL, Williams C, et al. Zika virus
infection among U.S. pregnant travelers — August 2015–February 2016. MMWR Morb Mortal Wkly Rep 2016;65:211-4.
19. Mlakar J, Korva M, Tul N, et al. Zika virus associated with
microcephaly. N Engl J Med 2016;374:951-8.
20. Schuler-Faccini L, Ribeiro EM, Feitosa IM, et al. Possible association between Zika virus infection and microcephaly — Brazil, 2015. MMWR Morb Mortal Wkly Rep 2016;65:59-62.
21. Ventura CV, Maia M, Bravo-Filho V, Góis AL, Belfort R Jr.
Zika virus in Brazil and macular atrophy in a child with microcephaly. Lancet 2016;387:228.
22. Miranda-Filho Dde B, Martelli CM, Ximenes RA, et al. Initial
description of the presumed congenital Zika syndrome. Am J
Public Health 2016;106:598-600.
23. Oliveira Melo AS, Malinger G, Ximenes R, Szejnfeld PO,
Alves Sampaio S, Bispo de Filippis AM. Zika virus intrauterine
infection causes fetal brain abnormality and microcephaly: tip
of the iceberg? Ultrasound Obstet Gynecol 2016;47:6-7.

24. Corona-Rivera JR, Corona-Rivera E, Romero-Velarde E,
Hernández-Rocha J, Bobadilla-Morales L, Corona-Rivera A. Report and review of the fetal brain disruption sequence. Eur J
Pediatr 2001;160:664-7.
25. National Birth Defects Prevention Network. Major birth defects data from population-based birth defects surveillance programs in the United States, 2006-2010. August 2013 (http://www
.nbdpn.org/docs/DataDirectory2013_NBDPN_AR.pdf).
26. Bale JF Jr. Fetal infections and brain development. Clin Perinatol 2009;36:639-53.
27. Dick GW. Zika virus. II. Pathogenicity and physical properties. Trans R Soc Trop Med Hyg 1952;46:521-34.
28. Bell TM, Field EJ, Narang HK. Zika virus infection of the
central nervous system of mice. Arch Gesamte Virusforsch 1971;
35:183-93.
29. Tang H, Hammack C, Ogden SC, et al. Zika virus infects
human cortical neural progenitors and attenuates their growth.
Cell Stem Cell 2016 March 4 (Epub ahead of print).
30. Public Affairs Committee of the Teratology Society. Causation in teratology-related litigation. Birth Defects Res A Clin Mol
Teratol 2005;73:421-3.
31. Teixeira MG, da Conceição N Costa M, de Oliveira WK,
Nunes ML, Rodrigues LC. The epidemic of Zika virus-related
microcephaly in Brazil: detection, control, etiology, and future
scenarios. Am J Public Health 2016;106:601-5.
32. Reefhuis J, Gilboa SM, Johansson MA, et al. Projecting
month of birth for at-risk infants after Zika virus disease outbreaks. Emerg Infect Dis (in press).
33. Mitchell AA. Proton-pump inhibitors and birth defects —
some reassurance, but more needed. N Engl J Med 2010;363:
2161-3.
34. Costa F, Sarno M, Khouri R, et al. Emergence of congenital
Zika syndrome: viewpoint from the front lines. Ann Intern Med
2016 February 24 (Epub ahead of print).
35. Chan JF, Choi GK, Yip CC, Cheng VC, Yuen KY. Zika fever
and congenital Zika syndrome: an unexpected emerging arboviral disease. J Infect 2016 March 3 (Epub ahead of print).
36. Abuelo D. Microcephaly syndromes. Semin Pediatr Neurol
2007;14:118-27.
37. Byrd AL, Segre JA. Infectious disease: adapting Koch’s postulates. Science 2016;351:224-6.
38. Fredricks DN, Relman DA. Sequence-based identification of
microbial pathogens: a reconsideration of Koch’s postulates.
Clin Microbiol Rev 1996;9:18-33.
39. Williams JV. Déjà vu all over again: Koch’s postulates and
virology in the 21st century. J Infect Dis 2010;201:1611-4.
40. Hill AB. The environment and disease: association or causation? Proc R Soc Med 1965;58:295-300.
41. Frank C, Faber M, Stark K. Causal or not: applying the Bradford Hill aspects of evidence to the association between Zika
virus and microcephaly. EMBO Mol Med 2016 March 14 (Epub
ahead of print).
42. Instituto Nacional de Salud. Boletin Epidemiologico Semanal. April 2016 (http://www.ins.gov.co/boletin-epidemiologico/
Boletn%20Epidemiolgico/2016%20Boletín%20epidemiológico
%20semana%2012.pdf).
43. Hubálek Z, Rudolf I, Nowotny N. Arborviruses pathogenic
for domestic and wild animals. Adv Virus Res 2014;89:201-75.
44. Palacios R, Poland GA, Kalil J. Another emerging arbovirus,
another emerging vaccine: targeting Zika virus. Vaccine 2016
March 23 (Epub ahead of print).
DOI: 10.1056/NEJMsr1604338
Copyright © 2016 Massachusetts Medical Society.

n engl j med nejm.org

The

n e w e ng l a n d j o u r na l

m e dic i n e

Special Article

Declines in Unintended Pregnancy
in the United States, 2008–2011
Lawrence B. Finer, Ph.D., and Mia R. Zolna, M.P.H.

A BS T R AC T
BACKGROUND

The rate of unintended pregnancy in the United States increased slightly between
2001 and 2008 and is higher than that in many other industrialized countries.
National trends have not been reported since 2008.
METHODS

We calculated rates of pregnancy for the years 2008 and 2011 according to women’s and girls’ pregnancy intentions and the outcomes of those pregnancies. We
obtained data on pregnancy intentions from the National Survey of Family Growth
and a national survey of patients who had abortions, data on births from the National Center for Health Statistics, and data on induced abortions from a national
census of abortion providers; the number of miscarriages was estimated using
data from the National Survey of Family Growth.

From the Guttmacher Institute, New York.
Address reprint requests to Dr. Finer at
the Guttmacher Institute, 125 Maiden
Ln., 7th Fl., New York, NY 10038, or at
lfiner@guttmacher.org.
N Engl J Med 2016;374:843-52.
DOI: 10.1056/NEJMsa1506575
Copyright © 2016 Massachusetts Medical Society.

RESULTS

Less than half (45%) of pregnancies were unintended in 2011, as compared with
51% in 2008. The rate of unintended pregnancy among women and girls 15 to 44
years of age declined by 18%, from 54 per 1000 in 2008 to 45 per 1000 in 2011.
Rates of unintended pregnancy among those who were below the federal poverty
level or cohabiting were two to three times the national average. Across population
subgroups, disparities in the rates of unintended pregnancy persisted but narrowed between 2008 and 2011; the incidence of unintended pregnancy declined by
more than 25% among girls who were 15 to 17 years of age, women who were
cohabiting, those whose incomes were between 100% and 199% of the federal
poverty level, those who did not have a high school education, and Hispanics. The
percentage of unintended pregnancies that ended in abortion remained stable during the period studied (40% in 2008 and 42% in 2011). Among women and girls
15 to 44 years of age, the rate of unintended pregnancies that ended in birth declined from 27 per 1000 in 2008 to 22 per 1000 in 2011.
CONCLUSIONS

After a previous period of minimal change, the rate of unintended pregnancy in
the United States declined substantially between 2008 and 2011, but unintended
pregnancies remained most common among women and girls who were poor and
those who were cohabiting. (Funded by the Susan Thompson Buffett Foundation
and the National Institutes of Health.)

n engl j med 374;9 nejm.org March 3, 2016

The New England Journal of Medicine
Downloaded from nejm.org at Stephen B. Thacker CDC Library on September 7, 2017. For personal use only. No other uses without permission.
Copyright © 2016 Massachusetts Medical Society. All rights reserved.

843

The

n e w e ng l a n d j o u r na l

he rate of unintended pregnancy
in a population is a central measure of
reproductive health; it indicates the extent
to which women and couples can determine
freely whether and when they have children. In
addition to supporting individual autonomy,
there is also a clear public health justification
for reducing the rate of unplanned pregnancy:
women and girls who have unintended pregnancies that result in births are more likely than
those who intended to become pregnant to have
inadequate or a delayed initiation of prenatal
care, to smoke and drink during pregnancy, and
to have premature and low-birth-weight infants;
they are also less likely to breast-feed. Increased
risks of physical and mental health problems
have also been reported in children of women
who have unplanned pregnancies.1-9 Many U.S.
policies and programs have recognized these
relationships and focus on reducing the rate of
unintended pregnancy and associated adverse
health outcomes.10-12
Although the rate of unintended pregnancy in
the United States decreased between the late
1980s and the mid-1990s,13 it plateaued by 200114
and increased slightly between 2001 and 2008,
the most recent year for which estimates are
available.15 The rate of unintended pregnancy in
the United States is substantially higher than
that in other highly industrialized regions such
as Western Europe.16 We used U.S. data on pregnancy intentions, released in December 2014 by
the National Center for Health Statistics (NCHS),
to calculate the incidence of unintended pregnancy in 2011.

Me thods
Study Design and Key Measures

The methods we used for this analysis are similar to those used in previously published studies.15,17 Among all U.S. females and key population subgroups, we determined the total number
of pregnancies that ended in birth, miscarriage
(i.e., fetal loss or stillbirth), and induced abortion and calculated the percentages of each of
these pregnancy outcomes that were unintended; we then divided the total number of unintended pregnancies by the population of women
and girls 15 to 44 years of age to obtain a rate
of unintended pregnancy per 1000 in this age
group.
844

m e dic i n e

Data Sources and Definitions

The numbers of U.S. births, miscarriages, and
abortions reported or estimated in 2011 and
2008 were derived from several sources. The
numbers of births were obtained from NCHS,18,19
which tabulates data from birth certificates to
obtain birth counts at the national level. Because
there is no recognized best estimate of the number — or method to obtain the number — of
miscarriages in a given year, we followed a procedure that was established by researchers at
NCHS20 using that center’s National Survey of
Family Growth (NSFG), a nationally representative in-home survey that collects information on
pregnancy and childbearing: we calculated the
ratio of miscarriages to births that were reported in the NSFG and multiplied that ratio by the
actual number of U.S. births to obtain our estimates of the number of miscarriages. The total
number of abortions, including both surgical and
medication abortions, for each year was obtained from a periodic census of all known abortion providers that was conducted by the Gutt
macher Institute.21 This census is considered to
be the most comprehensive source of data on the
incidence of abortion in the United States.22
Pregnancy intention was defined according to
a respondent’s answers to a series of retrospective survey questions about her desire to become
pregnant right before each pregnancy occurred.
If she reported that she did not want to become
pregnant at the time the pregnancy occurred,
but wanted to become pregnant in the future,
the pregnancy was categorized as mistimed. If a
respondent reported that she did not want to
become pregnant then or at any time in the future, the pregnancy was categorized as unwanted. We classified a pregnancy as unintended if it
was either mistimed or unwanted; an intended
pregnancy was one that was desired at the time
it occurred or sooner.
Data on pregnancy intentions (often called
intendedness) were obtained from two nationally representative sources. The percentages of
births and miscarriages that resulted from unintended pregnancies were calculated from the
2011–2013 NSFG. We evaluated 1975 pregnancies that ended between 2009 and 2013 (with
2011 as the central or reference year), as reported by the respondents; a respondent could
report more than one pregnancy. The percentages of abortions that followed unintended con-

n engl j med 374;9 nejm.org March 3, 2016

Unintended Pregnancy in the United States, 2008–2011

ceptions were calculated from the 2008 Abortion
Patient Survey that was conducted by the Guttmacher Institute.23 This paper-and-pencil survey
gathered information from a representative sample of 9493 women who had abortions in the
United States and is the most recent data set
available of its kind. The questions about pregnancy intention in the Abortion Patient Survey
were modeled on those in the NSFG. For both
data sets, the pregnancy outcomes were weighted to represent all pregnancies in the United
States in 2011.
Statistical Analysis

The percentages of births, miscarriages, and
abortions that resulted from unintended pregnancies were applied to the counts of each respective pregnancy outcome and then summed
to determine the total number of unintended
pregnancies. To calculate rates, we obtained
population counts according to age and according to race and ethnic group from the U.S. Census Bureau.24 All other distributions of population subgroups were derived from the Annual
Social and Economic Supplements of the U.S.
Census Bureau’s Current Population Survey,25 except for religious affiliation, which was derived
from the NSFG. Poor females were defined as
those with incomes below 100% of the federal
poverty level, and low-income females were those
whose incomes were between 100% and 199%
of the federal poverty level.
When calculating the percentage of unintended pregnancies that ended in abortion, we
excluded miscarriages in order to assess only
pregnancies in which the outcome was determined by the respondent. The rates of unintended pregnancy according to educational attainment were limited to women 20 years of age
or older; this age cutoff excluded most females
who had not yet completed schooling, yet still
included young women, who have had historically high rates of unintended pregnancy. We
also updated the rates of unintended pregnancy
for 1981, 1987, 2001, and 2008 — years that the
NSFG was fielded — to take into account updated population estimates and recent improvements in our analytic approach. Data on pregnancy intendedness were also collected in the
1995 survey of the NSFG but were excluded owing to concerns about the accuracy of the pregnancy intendedness data from that year.26

We performed analyses at an aggregate level
and separately for each population subgroup: we
combined data on pregnancy intention, pregnancy outcomes, and populations from several
different sources to calculate rates, which made
it difficult to assess the reliability of our estimates and of the change over time. Because most
of the uncertainty around the rate estimates was
attributable to the percentage of pregnancies
that were unintended (since the numbers of
pregnancies and population denominators are
based largely on generally complete census data),
we performed a supplementary analysis to calculate 95% confidence intervals for the percentage
of pregnancies that were unintended using a
merged data set that combined the sample of
births and miscarriages from the NSFG with the
sample of abortions from the Abortion Patient
Survey. We then used this range of percentages
to calculate the 95% confidence intervals around
the rate estimates. Although these percentages
are expected to be less accurate than the ones
calculated in the aggregate manner, the 95%
confidence intervals around these percentages
should represent the variance around the rate
estimates.
The above approach uses two different data
sources for pregnancy intention. We also used a
single data set, the NSFG, to calculate a test
statistic for the change between 2008 and 2011
in the percentage of pregnancies that were unintended. Using the NSFG alone for all pregnancy
outcomes allows for a simple calculation of the
test statistic. Abortions are underreported in the
NSFG, and therefore the percentages calculated
using this approach were expected to be lower
than those in our main analysis. Nonetheless,
we considered this analysis of trends to be reasonable, because the underreporting of abortions has not changed substantially over time.27,28

R e sult s
Findings at the National Level

In 2011, a total of 6.1 million pregnancies occurred in the United States (Table 1); 45% of
these pregnancies (2.8 million) were unintended, as compared with 51% of the pregnancies in
2008. There were 45 unintended pregnancies for
every 1000 women and girls 15 to 44 years of
age in 2011, as compared with a rate of 54 per
1000 women and girls 15 to 44 years of age in

n engl j med 374;9 nejm.org March 3, 2016

845

846
328

967

831

1494

2635

≥200%

1785
1595

High school graduate or GED equivalent

Some college or associate’s degree

College graduate
3190
1101

White non-Hispanic

Black non-Hispanic

Race and ethnic group¶

813
1358

Not a high school graduate

Educational attainment§

1373

100–199%

699

1201

428

813

738

363

784

709

1286

2011

132

105

116

187

152

209

320

119

141

168

163

115

106

2008

Total

122

101

109

162

111

184

254

121

141

157

138

2011

101

137

198

104

2008

112

141

2011

Unintended

Pregnancy Rate†

n engl j med 374;9 nejm.org March 3, 2016

−15

−13

−14

−16

−2

−28

−20

−32

−18

−29

−16

−21

−15

−12

−13

−22

−20

−44

−28

−18

2131

2008

Percentage of Pregnancies
That Were Unintended

n e w e ng l a n d j o u r na l

<100%

Income as a percentage of the federal poverty
level

Cohabiting

262

378

Formerly married, not cohabiting

954

1181

Never married, not cohabiting

731

3084

Currently married

Relationship status

≥35 yr

444

1440

691

1650

30–34 yr

878

305

25–29 yr

402
1494

18–19 yr

124

430

2779

Unintended

20–24 yr

173

574

6138

Total

No. of Pregnancies, 2011
(in Thousands)

15–17 yr

15–19 yr

Age group‡

All females

Characteristic

% Change in Rate
of Unintended
Pregnancy,
2008 to 2011

Table 1. Number of Pregnancies, Percentage That Were Unintended, Pregnancy Rates, and Percentage Change in the Rate of Unintended Pregnancies among All U.S. Females, 2008
and 2011.*

The

m e dic i n e

2008, which corresponds to an 18% decline over
this period (Table 1). This was the first substantial decline since at least 1981 (Fig. 1). The rate
of intended pregnancy increased slightly from
51 to 53 per 1000 women and girls 15 to 44 years
of age (data not shown); as a result, the overall
rate of pregnancy decreased from 106 to 98 per
1000 women and girls 15 to 44 years of age.
In 2011, the percentage of unintended pregnancies (excluding miscarriages) that ended in
abortion was 42% (Table 2). This percentage
changed little from 2008, when it was 40%. The
rate of births that resulted from unintended
pregnancies declined from 27 to 22 per 1000
women and girls 15 to 44 years of age during
the period studied.
Findings for Population Subgroups

The decline in rates of unintended pregnancy
was seen in almost every demographic group we
examined (Table 1). For example, the rate declined in every age group. However, the highest
rate of unintended pregnancy in 2011 was seen
among women 20 to 24 years of age, followed by
women 18 to 19 and women 25 to 29 years of
age. The percentage of unintended pregnancies
that ended in abortion did not vary substantially
according to age group, although the percentage
increased between 2008 and 2011 among girls
15 to 17 years of age; as a result, the pattern of
births that resulted from unintended pregnancies
reflected that of unintended pregnancy, with the
highest rates observed among women 18 to 29
years of age and declines in every age group.
The rate of unintended pregnancy varied ac-

Unintended Pregnancy Rate
(no. per 1000)

70
60
50
40
30
20
10
20
08
20
11

20
01

87
19

0
19

68
101
113
49
639
None

1311

109
46

34
44
598

49
661

205

1427

Other

Catholic

106

101
39

571
1473

703
1329

Evangelical Protestant

Mainline Protestant

1274
Protestant

Religious affiliation

2803

103
45
50

2008

140
50

2011

691
1387

2008
Unintended
Total

Hispanic

Characteristic

Percentage of Pregnancies
That Were Unintended
No. of Pregnancies, 2011
(in Thousands)

* Numbers may not sum to group totals because of rounding. GED denotes General Educational Development.
† Rates are reported as the number of pregnancies per 1000 women and girls 15 to 44 years of age.
‡ Girls younger than 15 years of age were excluded because of insufficient data. For the category 35 years of age or older, the numerator is the number of pregnancies among women 35
years of age or older and the population denominator is the number of women 35 to 44 years of age.
§ Calculations by educational attainment were limited to women 20 years of age or older.
¶ Race and ethnic group were self-reported. Data from women and girls who reported their race or ethnic group as “other” are not included here.

−26

42
109

−10

−11
54
102

48
91

−11

−27

57
95

52
93

−19
43

−26
58
79
116

2011
2008
2011

Unintended
Total

Pregnancy Rate†

of Unintended
Pregnancy,
2008 to 2011

Unintended Pregnancy in the United States, 2008–2011

Figure 1. Rates of Unintended Pregnancy, 1981–2011.
Rates are reported as the number of unintended pregnancies per 1000 women and girls 15 to 44 years of age.

n engl j med 374;9 nejm.org March 3, 2016

847

The

n e w e ng l a n d j o u r na l

m e dic i n e

Table 2. Percentage of Unintended Pregnancies That Ended in Abortion and Rate of Unintended Pregnancies That
Ended in Birth for All U.S. Females, 2008 and 2011.
Percentage of Unintended
Pregnancies That Ended
in Abortion*

Characteristic
All females

Rate of Unintended
Pregnancies That Ended
in Birth†

2008

2011

2008

2011

Age group‡
15–19 yr

15–17 yr

18–19 yr

20–24 yr

25–29 yr

30–34 yr

≥35 yr

Currently married

Never married, not cohabiting

Formerly married, not cohabiting

Cohabiting

101

Relationship status

Income as a percentage of the federal poverty level
<100%

100–199%

≥200%

Not a high school graduate

High school graduate or GED equivalent

Some college or associate’s degree

College graduate

Educational attainment§

Race and ethnic group¶
White non-Hispanic

Black non-Hispanic

Hispanic

Religious affiliation
Protestant

Mainline Protestant

Evangelical Protestant

Catholic

Other

None

* Pregnancies that ended in miscarriage were excluded.
† Rates are reported as the number of unintended pregnancies per 1000 women and girls 15 to 44 years of age.
‡ Girls younger than 15 years of age were excluded because of insufficient data. For the category 35 years of age or older,
the numerator is the number of pregnancies among women 35 years of age or older and the population denominator is the
number of women 35 to 44 years of age.
§ Calculations by educational attainment were limited to women 20 years of age or older.
¶ Race and ethnic group were self-reported. Data from women and girls who reported their race or ethnic group as “other”
are not included here.

848

n engl j med 374;9 nejm.org March 3, 2016

Unintended Pregnancy in the United States, 2008–2011

White non-Hispanic

Black non-Hispanic

Hispanic

140

Unintended Pregnancy Rate
(no. per 1000)

cording to relationship status. Women who were
married had the lowest rate of unintended pregnancy in 2011; by contrast, the rate among those
who were unmarried but cohabiting was more
than quadruple that among those who were married. However, the rate declined sharply between
2008 and 2011 among women who were cohabiting and to a lesser extent among those who
were married or never married; those who were
formerly married were the only group that had
an increase in the rate of unintended pregnancy
between 2008 and 2011. When an unintended
pregnancy occurred, women who were married
were much less likely to have an abortion than
were those who were unmarried.
We found a strong inverse association between both income level and educational attainment and the rate of unintended pregnancy.
However, the rate of unintended pregnancy declined between 2008 and 2011 in every income
and education group, with the largest declines
occurring among poor females and those who
did not have a high school education. As a result, the absolute differences by income and
education narrowed between 2008 and 2011. In
addition to having higher rates of unintended
pregnancy, poor and less-educated females were
less likely to have induced abortions to end unintended pregnancies; as a result, the income
and education disparities in the rate of unintended pregnancies that ended in birth were
even greater than the disparities in the unintended pregnancy rate. Nevertheless, the rate of
births that resulted from unintended pregnancies declined in virtually every income and education group.
There were substantial disparities in the rates
of unintended pregnancy in 2011 according to
race and ethnic group, even after income was
accounted for (Fig. 2). However, the rate of unintended pregnancy declined between 2008 and
2011 in all racial and ethnic groups, with the
largest decline among Hispanics. In 2011, the
percentage of unintended pregnancies that ended in abortion was highest among blacks, and
the rate of birth resulting from unintended pregnancies was lower among whites than among
both blacks and Hispanics.
The rates of unintended pregnancy and of
births resulting from unintended pregnancies
also declined between 2008 and 2011 among
women and girls of every religious affiliation
assessed. In both years, these rates were highest

120
100
80
60
40
20
0
<100%

100– 199%

≥200%

Income as a Percentage of the Federal Poverty Level

Figure 2. Rates of Unintended Pregnancy According to Income and Race
and Ethnic Group, 2011.
Rates are reported as the number of unintended pregnancies per 1000
women and girls 15 to 44 years of age.

among mainline Protestants and among those
with no religious affiliation.
Figure 3 shows that there have been declines
in rates of unintended pregnancy in the most recent period across all strata of age, income, and
race and ethnicity; this represents a change in the
overall pattern since 1981. The greatest reductions
were noted among women 20 to 24 years of age,
poor and low-income women and girls, and Hispanics.
Supplementary Analysis

In the supplementary analysis to assess the variance around our estimates (Table S1 in the
Supplementary Appendix, available with the full
text of this article at NEJM.org), we found a decline in the percentage of reported pregnancies
that were unintended, from 46% in 2008 to 39%
in 2011 (P = 0.01). Similarly, the supplementary
analysis yielded a point estimate and a 95% confidence interval for the rate of unintended pregnancies of 45 (95% confidence interval [CI], 41
to 49) per 1000 women and girls 15 to 44 years
of age in 2011, as compared with a rate of 54
(95% CI, 51 to 58) per 1000 women and girls 15
to 44 years of age in 2008. The confidence intervals do not overlap, which corroborates the finding of a decline.
Population subgroups with larger point esti-

n engl j med 374;9 nejm.org March 3, 2016

849

The

n e w e ng l a n d j o u r na l

Unintended Pregnancy Rate
(no. per 1000)

140
120
100
20–24 yr
18–19 yr
25–29 yr
All females

80
60
40

30–34 yr
15–17 yr
≥35 yr

01
20

87
19

Unintended Pregnancy Rate
(no. per 1000)

B Income as a Percentage of the Federal Poverty Level
140
120

<100%

100
80
60
40

100–199%
All females

≥200%

01
20

87
19

C Race and Ethnic Group
Unintended Pregnancy Rate
(no. per 1000)

m e dic i n e

Discussion

A Age

140
120
100
80

Black non-Hispanic

Hispanic
All females
White non-Hispanic

40
20

20
08
20
11

20
01

19
87

19
81

Figure 3. Rates of Unintended Pregnancy According to Key Sociodemographic
Characteristics, 1981–2011.
Rates are reported as the number of unintended pregnancies per 1000 women
and girls 15 to 44 years of age.

mates for the rate of unintended pregnancy
generally had wider 95% confidence intervals.
The results of the supplementary analysis supported the finding of differences in rates of
unintended pregnancy across strata of age, relationship status, income, education, and race and
ethnicity; the results did not support a finding
of clear differences in the rates across strata of
religious affiliation.
850

After a long period of minimal change, the rate
of unintended pregnancy in the United States
declined substantially between 2008 and 2011.
The rate of 45 unintended pregnancies per 1000
in 2011 was the lowest level seen in at least three
decades. The decline occurred in nearly all demographic groups, including those defined by
age, income, education, race and ethnicity, and
religious affiliation.
The decline we observed corroborates the
findings of a recent study29 that examined rates
of unintended pregnancy at the state level; this
study used a different source for girls’ and
women’s reports of pregnancy intention — the
Pregnancy Risk Assessment Monitoring System
of the Centers for Disease Control and Prevention — to produce state-specific estimates. In
that study, declines of 5% or more between 2006
and 2010 occurred in 28 of 41 states that had
data for both years.
Our analysis did not address factors that
might explain the decline between 2008 and
2011, but several possible factors should be considered. Changes in sexual behavior are unlikely
to have been a major driver. The incidence of
sexual activity tends not to change much among
adults,30 and among women 18 to 19 years of
age, the decline in the rate of unintended pregnancy occurred despite virtually no change over
the course of the period studied in the percentage who reported ever having sex31; because
younger teens have relatively few pregnancies,
any change in their behavior would have relatively little effect on the overall rate of unintended pregnancy. Changes in the composition
of the population are also not likely to explain
the decline in the rate of unintended pregnancy;
in fact, there is evidence that the percentage of
the population composed of women and girls
with higher rates of unintended pregnancy, such
as those who were poor or Hispanic, increased
over time,24,25,32 and the decline in the rate of
unintended pregnancy occurred despite this increase.
Change in the desire for pregnancy may have
contributed to the decline in the rate of unintended pregnancies. Surveys of women in 2009
during the recession indicated that many women
intended to reduce or delay their childbearing
because of changing economic conditions.33 As
Americans recovered from the recession, it is

n engl j med 374;9 nejm.org March 3, 2016

Unintended Pregnancy in the United States, 2008–2011

possible that there was a corresponding increase
in desired pregnancy, which would have led to a
shift away from unplanned pregnancies; our
analyses show that there was a small increase in
the rate of intended pregnancy between 2008
and 2011.
A likely explanation for the decline in the rate
of unintended pregnancy is a change in the frequency and type of contraceptive use over time.
Evidence shows that the overall use of any
method of contraception among women and
girls at risk for unintended pregnancy increased
slightly between 2008 and 2012.34,35 More important, the use of highly effective long-acting
methods, particularly intrauterine devices, among
U.S. females who used contraception increased
from 4% to 12% between 2007 and 2012,36 and
this increase occurred in almost all demographic groups.37,38 In a 2012 study, women and girls
at high risk of unintended pregnancy who had
free access to and used highly effective methods
of contraception had much lower rates of unintended pregnancy than did those who used
other methods, including commonly used methods such as the oral contraceptive pill.39
Although the differences in rates of unintended pregnancy across demographic groups
narrowed over time, large disparities were still
present in 2011. In particular, poor, black, and
Hispanic women and girls continued to have
much higher rates of unintended pregnancy
than did whites and those with higher incomes.
Much more progress can be made in eliminating
these disparities. The rate of unintended pregnancy in Western Europe is 40% lower than the
rate in the United States,16 and the rate associated with higher incomes in the United States is
similar to the rate among all women in Western
Europe.
The observed decrease in the rate of unintended pregnancy preceded the implementation
of several provisions in the Affordable Care Act
that should improve coverage for contraceptive
services, including the option for young people
References
1. Kost K, Lindberg L. Pregnancy intentions, maternal behaviors, and infant
health: investigating relationships with
new measures and propensity score analysis. Demography 2015;52:83-111.
2. Axinn WG, Barber JS, Thornton A.
The long-term impact of parents’ childbearing decisions on children’s self-esteem.
Demography 1998;35:435-43.

up to 26 years of age to remain on their parents’
health insurance plans and a provision that requires insurance plans to cover contraception at
no out-of-pocket cost. If these provisions lead to
greater use of contraception overall or to increased use of highly effective methods among
those who want them, the rate of unintended
pregnancy could continue to decline.
A limitation of our study is that we used
socioeconomic and other demographic information on women and girls from the 2008
Abortion Patient Survey to estimate both the
2008 and 2011 counts of women and girls who
had abortions by characteristic. These counts
might have changed through 2011. For example,
the percentage of abortion patients who were
poor increased from 2000 to 2008,23 and it is
possible that this percentage continued to increase from 2008 to 2011. If an increase in this
percentage occurred from 2008 to 2011, the
number of poor women and girls who had an
unintended pregnancy in 2011, as well as the
rate of unintended pregnancy, could have been
underestimated; thus, the decline in the rate of
unintended pregnancy among poor women and
girls would be overestimated, and the decline in
the rate of unintended pregnancy among those
with higher incomes would be underestimated.
Our findings show a substantial decline in
the rate of unintended pregnancy in the United
States between 2008 and 2011, to a historic low.
Nonetheless, nearly half of all pregnancies in
2011 were still unintended, and major disparities remained among women and girls according to socioeconomic status and race and ethnic
group.
Supported by a grant from the Susan Thompson Buffett Foundation. Additional support was provided by the Guttmacher
Center for Population Research Innovation and Dissemination
through grant 5 R24 HD074034 from the National Institutes of
Health.
Disclosure forms provided by the authors are available with
the full text of this article at NEJM.org.
We thank Heather Boonstra, M.A., Rachel Jones, Ph.D., and
Kathryn Kost, Ph.D., for their comments on earlier versions of
the manuscript.

3. Barber JS, Axinn WG, Thornton A.

Unwanted childbearing, health, and
mother-child relationships. J Health Soc
Behav 1999;40:231-57.
4. Dott M, Rasmussen SA, Hogue CJ,
Reefhuis J. Association between pregnancy
intention and reproductive-health related
behaviors before and after pregnancy
recognition, National Birth Defects Pre-

vention Study, 1997-2002. Matern Child
Health J 2010;14:373-81.
5. Dye TD, Wojtowycz MA, Aubry RH,
Quade J, Kilburn H. Unintended pregnancy and breast-feeding behavior. Am J
Public Health 1997;87:1709-11.
6. Hellerstedt WL, Pirie PL, Lando HA,
et al. Differences in preconceptional and
prenatal behaviors in women with intend-

n engl j med 374;9 nejm.org March 3, 2016

851

Unintended Pregnancy in the United States, 2008–2011

ed and unintended pregnancies. Am J
Public Health 1998;88:663-6.
7. Mayer JP. Unintended childbearing,
maternal beliefs, and delay of prenatal
care. Birth 1997;24:247-52.
8. Orr ST, Miller CA, James SA, Babones
S. Unintended pregnancy and preterm
birth. Paediatr Perinat Epidemiol 2000;14:
309-13.
9. Taylor JS, Cabral HJ. Are women with
an unintended pregnancy less likely to
breastfeed? J Fam Pract 2002;51:431-6.
10. Institute of Medicine. Clinical preventive services for women: closing the gaps.
Washington, DC:National Academies
Press, 2011.
11. Healthy People 2020 topics and objectives. Washington, DC:Department
of Health and Human Services, 2010
(http://www.healthypeople.gov/2020/topics
-objectives/topic/family-planning/objectives).
12. The Surgeon General’s call to action to
support breastfeeding. Washington, DC:
Department of Health and Human Services,
Office of the Surgeon General, 2011.
13. Henshaw SK. Unintended pregnancy
in the United States. Fam Plann Perspect
1998;30:24–29, 46.
14. Finer LB, Henshaw SK. Disparities in
rates of unintended pregnancy in the
United States, 1994 and 2001. Perspect
Sex Reprod Health 2006;38:90-6.
15. Finer LB, Zolna MR. Shifts in intended and unintended pregnancies in the
United States, 2001-2008. Am J Public
Health 2014;104:Suppl 1:S43-8.
16. Sedgh G, Singh S, Hussain R. Intended and unintended pregnancies worldwide in 2012 and recent trends. Stud Fam
Plann 2014;45:301-14.
17. Finer LB, Zolna MR. Unintended
pregnancy in the United States: incidence
and disparities, 2006. Contraception 2011;
84:478-85.
18. Martin JA, Hamilton BE, Sutton PD,
Ventura SJ, Mathews TJ, Osterman MJ.
Births: final data for 2008. Natl Vital Stat
Rep 2010;59:1, 3-71.
19. Martin JA, Hamilton BE, Ventura SJ,
Osterman MJ, Mathews TJ. Births: final
data for 2011. Natl Vital Stat Rep 2013;62:
1-69, 72.
20. Ventura SJ, Curtin SC, Abma JC, Henshaw SK. Estimated pregnancy rates and
rates of pregnancy outcomes for the Unit-

ed States, 1990-2008. Natl Vital Stat Rep States: data from the 2006–2008 National
2012;60:1-21.
Survey of Family Growth. Hyattsville,
21. Jones RK, Jerman J. Abortion inci- MD:National Center for Health Statistics,
dence and service availability in the United 2011.
States, 2011. Perspect Sex Reprod Health 31. National Center for HIV/AIDS. Trends
2014;46:3-14.
in the prevalence of sexual behaviors and
22. Pazol K, Creanga AA, Burley KD, Jamie- HIV testing, national YRBS:1991–2013.
son DJ. Abortion surveillance — United Atlanta:Centers for Disease Control and
States, 2011. MMWR Surveill Summ 2014; Prevention, 2014 (http://www.cdc.gov/
63(11):1-41.
healthyyouth/d ata/y rbs/pdf/t rends/us_
23. Jones RK, Finer LB, Singh S. Charac- sexual_trend_yrbs.pdf).
teristics of U.S. abortion patients, 2008. 32. Current Population Survey, 2009 AnNew York:Guttmacher Institute, 2010 nual Social and Economic (ASEC) Supple(http://www.guttmacher.org/pubs/US
ment [machine-readable data file]/conduct-Abortion-Patients.pdf).
ed by the Bureau of the Census for the
24.
CDC WONDER online database: Bureau of Labor Statistics. Washington, DC:
bridged-race population estimates, United U.S. Census Bureau, 2009 (https://w ww
States July 1st resident population by .census.gov/prod/t echdoc/cps/cpsmar09
state, county, age, sex, bridged-race, and .pdf ).
Hispanic origin, compiled from 2000– 33. A real-time look at the impact of the
2013 bridged-race population estimates recession on women’s family planning
(Vintage 2013). Department of Health and and pregnancy decisions. New York:GuttHuman Services, National Center for macher Institute, 2009.
Health Statistics, Division of Vital Statistics 34. Daniels K, Daugherty J, Jones J. Cur(http://wonder.cdc.gov/bridged-race-v2013 rent contraceptive status among women
.html).
aged 15–44: United States, 2011–2013.
25. Current Population Survey, 2012 An- Washington, DC:Department of Health
nual Social and Economic (ASEC) Supple- and Human Services, 2014.
ment [machine-readable data file]/conduct- 35. Jones J, Mosher WD, Daniels K. Current
ed by the Bureau of the Census for the contraceptive use in the United States,
Bureau of Labor Statistics. Washington, 2006–2010, and changes in patterns of
DC:U.S. Census Bureau, 2012 (https://w ww use since 1995. Hyattsville, MD:National
.census.gov/prod/t echdoc/cps/cpsmar12
Center for Health Statistics, 2012.
.pdf ).
36. Use of highly effective contraceptives
26. Mosher WD, Jones J, Abma JC. Intend- in the U.S. continues to rise, with likely
ed and unintended births in the United implications for declines in unintended
States: 1982–2010. Hyattsville, MD:Na- pregnancy and abortion. New York:Gutttional Center of Health Statistics, 2012.
macher Institute, 2014 (http://www.gutt27. Fu H, Darroch JE, Henshaw SK, Kolb E. macher.org/media/inthenews/2014/12/12/
Measuring the extent of abortion under- index.html).
reporting in the 1995 National Survey of 37. Branum A, Jones J. Trends in longFamily Growth. Fam Plann Perspect 1998; acting reversible contraception use among
30:128-33, 138.
U.S. women aged 15–44. Hyattsville, MD:
28. Jones RK, Kost K. Underreporting of National Center for Health Statistics,
induced and spontaneous abortion in the 2015.
United States: an analysis of the 2002 Na- 38. Kavanaugh ML, Jerman J, Finer LB.
tional Survey of Family Growth. Stud Fam Changes in use of long-acting reversible
Plann 2007;38:187-97.
contraceptive methods among United
29. Kost K. Unintended pregnancy rates States women, 2009–2012. Obstet Gyneat the state level: estimates for 2010 and col 2015;126:917-27. .
trends since 2002. New York:Guttmacher 39. Winner B, Peipert JF, Zhao Q, et al.
Institute, 2015.
Effectiveness of long-acting reversible
30. Chandra A, Mosher W, Copen C, Sio- contraception. N Engl J Med 2012;366:
nean C. Sexual behavior, sexual attrac- 1998-2007.
tion, and sexual identity in the United Copyright © 2016 Massachusetts Medical Society.

We are seeking a Deputy Editor to join our editorial team at the New England Journal
of Medicine. The Deputy Editor will review, select, and edit manuscripts for the
Journal as well as participate in planning the Journal’s content. He or she will also
ensure that the best available research is submitted to the Journal, identify potential
Journal contributors, and invite submissions as appropriate.
Applicants must have an M.D. degree and at least 5 years of research experience.
The successful candidate will be actively involved in patient care and will continue
that involvement (approximately 20% of time) while serving in this role.
To apply, visit https://mmscareers.silkroad.com.

852

n engl j med 374;9 nejm.org March 3, 2016

Articles

The Zika Contraception Access Network: a feasibility
programme to increase access to contraception in
Puerto Rico during the 2016–17 Zika virus outbreak
Eva Lathrop, Lisa Romero, Stacey Hurst, Nabal Bracero, Lauren B Zapata, Meghan T Frey, Maria I Rivera, Erin N Berry-Bibee, Margaret A Honein,
Judith Monroe, Denise J Jamieson

Summary

Background Prevention of unintended pregnancy is a primary strategy to reduce adverse pregnancy and birth
outcomes related to Zika virus infection. The Zika Contraception Access Network (Z-CAN) aimed to build a network
of health-care providers offering client-centred contraceptive counselling and the full range of reversible contraception
at no cost to women in Puerto Rico who chose to prevent pregnancy during the 2016–17 Zika virus outbreak. Here, we
describe the Z-CAN programme design, implementation activities, and baseline characteristics of the first
21 124 participants.
Methods Z-CAN was developed by establishing partnerships between federal agencies, territorial health agencies,
private corporations, and domestic philanthropic and non-profit organisations in the continental USA and Puerto
Rico. Private donations to the National Foundation for the Centers for Disease Control and Prevention (CDCF)
secured a supply of reversible contraceptive methods (including long-acting reversible contraception), made available
to non-sterilised women of reproductive age at no cost through provider reimbursements and infrastructure
supported by the CDCF. To build capacity in contraception service provision, doctors and clinic staff from all public
health regions and nearly all municipalities in Puerto Rico were recruited into the programme. All providers
completed 1 day of comprehensive training in contraception knowledge, counselling, and initiation and management,
including the insertion and removal of long-acting reversible contraceptives (LARCs). Z-CAN was announced through
health-care providers, word of mouth, and a health education campaign. Descriptive characteristics of programme
providers and participants were recorded, and we estimated the factors associated with choosing and receiving a
LARC method. As part of a Z-CAN programme monitoring plan, participants were invited to complete a patient
satisfaction survey about whether they had obtained free, same-day access to their chosen contraceptive method after
receiving comprehensive counselling, their perception of the quality of care they had received, and their satisfaction
with their chosen method and services.
Findings Between May 4, 2016, and Aug 15, 2017, 153 providers in the Z-CAN programme provided services to
21 124 women. 20 110 (95%) women received same-day provision of a reversible contraceptive method. Whereas only
767 (4%) women had used a LARC method before Z-CAN, 14 259 (68%) chose and received a LARC method at their
initial visit. Of the women who received a LARC method, 10 808 (76%) women had used no method or a least effective
method of contraception (ie, condoms or withdrawal) before their Z-CAN visit. Of the 3489 women who participated
in a patient satisfaction survey, 3068 (93%) of 3294 women were very satisfied with the services received, and
3216 (93%) of 3478 women reported receiving the method that they were most interested in after receiving counselling.
2382 (78%) of 3040 women rated their care as excellent or very good.
Interpretation Z-CAN was designed as a short-term response for rapid implementation of reversible contraceptive
services in a complex emergency setting in Puerto Rico and has served more than 21 000 women. This model could
be replicated or adapted as part of future emergency preparedness and response efforts.
Funding National Foundation for the Centers for Disease Control and Prevention.
Copyright © The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license.

Introduction
Prevention of unintended pregnancy is a primary strategy
to reduce adverse pregnancy and birth outcomes related
to Zika virus infection.1,2 Puerto Rico has the highest
number of symptomatic Zika virus infections in the USA
and US territories, including infections in women.3
Additionally, 65% of pregnancies in Puerto Rico are

Lancet Public Health 2018
Published Online
January 18, 2018
http://dx.doi.org/10.1016/
S2468-2667(18)30001-X
See Online/Comment
http://dx.doi.org/10.1016/
S2468-2667(18)30005-7
Department of Gynecology and
Obstetrics, Emory University,
Atlanta, GA, USA (E Lathrop MD,
E N Berry-Bibee MD,
D J Jamieson MD); Division of
Reproductive Health, National
Center for Chronic Disease
Prevention and Health
Promotion (L Romero DrPH,
S Hurst MPH, L B Zapata PhD)
and Division of Congenital and
Developmental Disorders,
National Center on Birth
Defects and Developmental
Disabilities (E Lathrop,
M T Frey MPH, M I Rivera MPH,
M A Honein PhD), Centers for
Disease Control and
Prevention, Atlanta, GA, USA;
University of Puerto Rico, Área
Centro Medico, San Juan,
Puerto Rico (N Bracero MD);
Puerto Rico Section of the
American College of
Obstetricians and
Gynecologists, San Juan,
Puerto Rico (N Bracero); Puerto
Rico Obstetrics and
Gynecology, San Juan,
Puerto Rico (N Bracero); and
National Foundation for the
Centers for Disease Control and
Prevention, Atlanta, GA, USA
(J Monroe MD)
Correspondence to:
Dr Eva Lathrop, Department of
Gynecology and Obstetrics,
Emory University, Atlanta, GA
30303, USA
[email protected]

unintended, and about 138 000 of the 715 000 women
aged 15–44 years in Puerto Rico are at risk for unintended
pregnancy.4 5–10% of the pregnancies with laboratoryconfirmed Zika virus infection that were reported to the
US Zika Pregnancy Registry resulted in a fetus or infant
with Zika-virus-associated birth defects, and the full
range of adverse development outcomes is not yet

www.thelancet.com/public-health Published online January 18, 2018 http://dx.doi.org/10.1016/S2468-2667(18)30001-X

Articles

Research in context
Evidence before this study
We searched PubMed for articles published on or before
April 1, 2016, using the terms “Contraceptive Choice Project”,
“Zika and family planning”, and “Zika and contraception”. The
Contraceptive CHOICE Project was a prospective cohort study of
10 000 women of reproductive age in St Louis, MO, USA, who
wanted to prevent pregnancy and initiate a new method of
contraception. The study was designed to introduce and
promote the use of long-acting reversible contraception (LARC)
methods, and the results showed that 65% of participating
women chose LARC methods when cost, provider, and facility
barriers were removed. In a report from April 1, 2016, early in
Puerto Rico’s 2016–17 Zika virus outbreak, women in the
country were shown to have a high unmet need for
contraception, high incidence of unintended pregnancy, poor
access to contraception, and the highest number of Zika
infections in the USA and US territories. We did not identify any
studies that described a contraception-focused programme as
part of the response to the Zika virus outbreak.

programme to rapidly expand access to contraceptives during a
major public health emergency response. The programme was
implemented quickly and was able to serve more women than
previous projects based on expansion of contraceptive access.
Z-CAN included introduction to and education about LARC
methods for both providers and patients with no previous
exposure to or experience with these newer contraceptive
methods.
Implications of all the available evidence
This large and rapidly established contraception programme
could be replicated in other areas with serious and complex
public health emergencies to ensure that unintended births are
averted. Although this programme was developed to prevent
unintended pregnancies and birth defects associated with Zika
virus infection, avoiding unintended pregnancy is an important
strategy for a wide variety of public health responses,
particularly in view of frequent disruptions in care and services
in emergency settings.

Added value of this study
The Zika Contraception Access Network (Z-CAN) is the first to
describe the large-scale implementation of a comprehensive

known.5 The threat of severe birth defects associated with
Zika virus infection during pregnancy underscores the
importance of contraception to prevent unintended
pregnancies. However, a review of existing data and
in-depth interviews with key informants early in the Zika
virus outbreak in March, 2016, demonstrated that
contraceptive access in Puerto Rico was limited by
reduced availability of the full range of reversible
methods, high out-of-pocket costs, insufficient provider
reimbursement, logistical barriers that limit same-day
provision, lack of patient education, and shortage of
providers trained in insertion, removal, and management
of long-acting reversible contraception (LARC), which
includes intrauterine devices and contraceptive implants.4
LARC is a highly effective, safe, cost-effective, and
user-friendly method of contraception that reduces
unintended pregnancy and abortion.6–9 In 2002–14, LARC
use in the USA increased from 2·4% to 14·3% of women
using contraception.10 However, LARC use in Puerto Rico
was low before the Zika virus outbreak, with estimates
indicating that less than 1% of women using
contraception used a LARC method.4
Recognising the importance of contraceptive access
during the Zika virus outbreak, the National Foundation
for the Centers for Disease Control and Prevention
(CDCF), with technical assistance from the Centers for
Disease Control and Prevention (CDC) and in
collaboration with a diverse group of stakeholders and
private donors, established the Zika Contraception
Access Network (Z-CAN) in Puerto Rico. Z-CAN was a
2

short-term response (from May, 2016, to September, 2017)
for rapid implementation of reversible contraceptive
services in a complex emergency setting. Z-CAN aimed
to build a network of health-care providers trained in
client-centred contraceptive counselling and same-day
provision of the full range of reversible contraceptive
methods (including LARC) at no cost to women who
choose to delay or avoid pregnancy, and to raise
awareness in women and families of contraception as a
primary prevention measure to reduce adverse pregnancy
and birth outcomes related to Zika virus infection. In
addition to access barriers, a history of coerced
sterilisation and concern for unethical testing of oral
contraceptives in Puerto Rico were important
considerations in programme design.11,12
Here we describe the Z-CAN programme design and
implementation activities and the baseline characteristics
of the first 21 124 women served through Z-CAN.

Methods

Programme design and implementation
Z-CAN was designed to address gaps in contraceptive
access and service provision in Puerto Rico as a preventive
measure to reduce the effect of Zika virus on infants. The
development of Z-CAN included several strategies to
rapidly reduce access barriers to contraception in
Puerto Rico’s health system, strengthen infrastructure to
support the Z-CAN programme, and work towards the
sustainability of reversible contraceptive services after
the Z-CAN programme ends (figure 1).

www.thelancet.com/public-health Published online January 18, 2018 http://dx.doi.org/10.1016/S2468-2667(18)30001-X

Articles

January 22
Activation of CDC
Emergency Operations
Center
Feb 23
Meeting with
CMS, HRSA, OPA
to discuss
contraception
access barriers in
Puerto Rico

March 7
FLASOG, ACOG,
SOGC statement on
Zika, including
contraception
March 23
CDC Public
Health Ethics
Unit provides
recommendations on
contraceptive
product
donations

April 20
CDCF identifies
pharmaceutical
distributor

April 1
CDC article on
contraceptive access
in Puerto Rico

May 23
First CDCF donor
agreement signed with
pharmaceutical
manufacturer for
contraceptive donation
June 15
PRDOH issues
Botiquin waivers
for Z-CAN
providers to stock
contraceptive
methods for
same-day
provision

May 4
Began Z-CAN
doctor and
clinic staff
proctoring

April 12
Z-CAN Key
Stakeholder
meeting

July 7
Puerto Rico public and
private health plans
leadership meeting to
discuss sustainability
post Z-CAN

Sept 19
Focus groups
conducted to test
health education
campaign
concepts in
Puerto Rico

Aug 10
Focus groups
conducted
among women
and men of
reproductive
age on
contraception
access in Puerto
Rico

Oct 26
Z-CAN doctor
(>97%) and
clinic staff
(>90%)
proctored

Dec 30
Premiered Ante
La Duda,
Pregunta Public
Service
Announcements

Nov 1
50% of Z-CAN
providers have
contraceptive
product

Dec 2
90% of Z-CAN
providers have
contraceptive
product

Nov 1
Z-CAN
webpage
with clinic
finder
launched

Dec 2
Ante La Duda, Pregunta
campaign website launched
with information on
contraceptive methods and
Z-CAN clinic finder

2016
Feb 25
Meeting with
ACOG office to
discuss
contraceptive
access in
Puerto Rico
Feb 22
Pregnancy and Birth
Defects Task Force
assess contraception
access in Puerto Rico

March 31
Z-CAN
programme
established

March 22
First
manufacturer
to offer
contraceptive
donation for
Zika response

Jan 16
5000 Z-CAN initial visits reported
to date

April 15
CDCF “Call to
Action” for
private sector
and philanthropy
to support Z-CAN
April 7
PRDOH issued AO
to broaden access
to effective
contraception for
Zika

May 19
Regulatory
and
distribution
plan finalised
between HHS
OIG, CMS,
FDA, and VA
April 30
Z-CAN programme
launched in Puerto
Rico with first
Z-CAN training

March 17
10 000 Z-CAN initial visits
reported to date

July 1
First CDCF
donor
agreement
signed with
philanthropic
foundation for
Z-CAN
June 2
Pharmaceutical
Distributor
agreement
signed

June 2
15 000 Z-CAN initial visits
reported to date

Aug 22
CDC non-research
determination
approval received
for 2-week
follow-up survey
with Z-CAN
patients

Oct 19
Z-CAN
2-week
follow-up
survey
launched

July 15
First order to
pharmaceutical distributor
submitted for first 20
approved Z-CAN providers
July 16
CDC, CDCF, CMS sustainability meeting with key
stakeholders

Nov 4
Ante La Duda, Pregunta (health
education campaign for women
of reproductive age in Puerto
Rico) Facebook Page launched

Sept 23
Z-CAN programme ends

2017
Jan 7
Community engagement and
outreach efforts began

March 1
Z-CAN meeting with CMS and
PRDOH to discuss sustainability

April 22
Z-CAN Mid-programme
Provider convening

June 3
CDC, CDCF, CMS, PRDOH
sustainability meeting

Aug 11
20 000 Z-CAN initial visits

Sept 29
CDC’s Emergency Operations Center
Zika response is deactivated

Figure 1: Zika Contraception Access Network (Z-CAN) major milestones, 2016–17
CDC=Centers for Disease Control and Prevention. CMS=Centers for Medicare and Medicaid Services. HRSA=Health Resources and Service Administration. OPA=Office of Population Affairs.
FLASOG=Federacion Latinoamericana de Sociedades de Obstetricia y Ginecologia. ACOG=American College of Obstetricians and Gynecologists. SOGC=The Society of Obstetricians and Gynecologists.
PRDOH=Puerto Rico Department of Health. AO=Administrative Order. HHS OIG=Health and Human Services Office of the Inspector General. FDA=US Food and Drug Administration. VA=Veterans
Administration.

The development of strong partnerships was crucial
in the design and implementation of Z-CAN. The
programme was built with a network of partners
including federal agencies, territorial health agencies,
private corporations, and domestic philanthropic and
non-profit organisations in the continental USA and
Puerto Rico. Private donors provided product
commitments to CDCF for the full range of reversible
contraceptive methods (including LARC methods).
CDCF established a plan for contraception procurement
and distribution adherent to US Food and Drug
Administration (FDA) and territorial guidelines and for
private donations through CDCF-supported provider
reimbursement and infrastructure costs to ensure
contraception was available to women at no cost.
The gaps in contraceptive access and service provision4
meant that it was necessary to build provider and staff
capacity in contraception knowledge, counselling, and
initiation and management, including the insertion and
removal of LARC. Z-CAN recruited doctors and clinic
staff (nurses and clinic administrators) from all public
health regions and nearly all municipalities on the island
who practised in private and publicly funded clinics and

who were interested in receiving training in the provision
of contraception.13 Doctors and clinic staff were not
recruited from municipalities with no community health
centres, government facilities, or private practices
providing women’s health care. Doctors and staff were
recruited through the Puerto Rico section of the American
College of Obstetricians and Gynecologists, Puerto Rico
Obstetrics and Gynecology, the Puerto Rico Department
of Health, the Puerto Rico Primary Care Association, the
Puerto Rico Health Insurance Administration, and
Medicaid-managed care organisations. Before Z-CAN,
none of the participating clinics routinely provided
levonorgestrel-releasing intrauterine devices or contraceptive implants, and access to copper intrauterine
devices was very limited. A 1-day comprehensive training
course offered participants an overview of Zika virus
(including the risk of sexual transmission and the
importance of condom use for disease prevention), a
tested curriculum on client-centred contraceptive
counselling, didactic information about the full range of
reversible contraceptives, a review of evidence-based
contraceptive guidelines,14,15 practical training in insertion
and removal of intrauterine devices (providers were

www.thelancet.com/public-health Published online January 18, 2018 http://dx.doi.org/10.1016/S2468-2667(18)30001-X

Articles

See Online for appendix

observed on three to five simulations),13 an FDA-approved
etonogestrel implant training, and a overview of Z-CAN
policies and procedures.
Provider reimbursement for these services was
previously identified as barriers to contraception access.4
Through Z-CAN, private donations were used to provide a
level of provider reimbursement that was commensurate
with Medicaid reimbursement rates in the continental
USA. This reimbursement covered client-centred
contraceptive counselling for women and their partners,
if desired, and method provision. If a LARC was provided,
the reimbursement fee was bundled to include both
insertion and removal at the time of the insertion visit to
ensure that women could have their LARC devices
removed when desired at no cost.
After initial training, a Z-CAN staff member and a
family planning specialist proctored providers and clinic
staff to ensure delivery of high-quality care. Proctoring
visits consisted of: direct observation of contraceptive
counselling, at least one insertion of an intrauterine
device, and staff interaction with patients; review of data
collection, inventory tracking, and billing procedures; and
a clinic audit to ensure that supplies, space, equipment,
and security were sufficient to participate in Z-CAN. If
provider, staff, and clinic met all readiness criteria, they
were authorised to receive contraceptive products and to
begin offering Z-CAN services.

Data collection and analysis
Women learned of Z-CAN through providers, word of
mouth, and a health education campaign involving
community engagement activities, Z-CAN materials,
posters in health centres, a campaign website, and a
Facebook page. Non-sterilised women of reproductive age
were eligible to receive Z-CAN services, irrespective of age
or insurance status. All Z-CAN services were provided
free of charge.
At the initial Z-CAN visit, women were assigned a
unique identification number. Providers and clinic staff
recorded women’s demographic information, reproductive
and contraception histories, and their chosen contraceptive
method. Data were submitted without personal identifying
information to the Z-CAN programme and entered into a
REDCap database hosted on a secure server.16
The data presented here are descriptive characteristics
of programme providers and women receiving Z-CAN
services. To examine factors associated with choosing and
receiving a LARC method, we estimated unadjusted and
adjusted prevalence ratios with 95% CI. Data were
analysed using SAS-callable SUDAAN version 11.0.0 to
account for clustering of patients within clinic-provider
dyads.
The CDC’s Public Health Ethics Committee (PHEC)
provided internal consultation during the programme
and project design to ensure no conflicts of interest
existed and to address any ethical concerns.17 The
Public Health Ethics Conflict of Interest Work Group,
4

part of the CDC Zika Response Emergency Operations
Center and comprised of individuals from the PHEC,
reviewed the Z-CAN programme proposal during its
design phase and recommended that the programme
offer the full range of reversible contraceptive methods
and have measures in place to prevent coercion of
women.
As part of the Z-CAN programme monitoring plan,
women were invited to participate in a 10 min
self-administered online survey within 2 weeks of their
initial visit. Z-CAN-trained clinic staff collected contact
information from women who did not opt out of being
contacted for future surveys. Women were invited to
participate in the survey via email or text message; those
without online access could complete the survey on the
telephone with programme staff. The survey measured
whether participants received free same-day access to the
contraceptive method of their choice after receiving
comprehensive counselling, patient perception of the
received quality of care, and satisfaction with their chosen
method and services. Perception of quality of care was
measured using the validated interpersonal quality of
family planning care scale,18 comprised of 11 items
measured using a five-point Likert scale (a score of 1 means
poor; a score of 5 means excellent; appendix). No personal
identifiers were collected, and unique identification
numbers were used to merge survey responses with initial
visit data. Women were considered non-respondents if
they did not complete the survey within 3 weeks after
confirmed receipt of email or text message invitation and
after up to three outreach attempts. Responses were
collected through Survey Monkey online software, and
respondents received a US$10 electronic gift card. We used
SAS version 9.3 to compare baseline characteristics of
survey respondents and non-respondents.
The Z-CAN programme and patient satisfaction
survey were determined by CDC to be non-research
public health practice activities and thus exempt from
Institutional Review Board review. The programme did
not obtain consent from women served by Z-CAN
providers. The women received a letter at their initial
visit that described the follow-up contact planned for
programme monitoring purposes and were given the
opportunity to opt out. Women who did not opt out
were invited to participate in the patient satisfaction
survey. If a woman chose to participate in the survey,
she did so by consenting to the survey within the online
environment.

Role of the funding source
The philanthropic donors to CDCF had no role in
programme design, data collection, data analysis, data
interpretation, or writing of the report. CDC provided
technical assistance in collaboration with CDCF for
programme design and implementation. The corresponding author had full access to all of the data and
the final responsibility to submit for publication.

www.thelancet.com/public-health Published online January 18, 2018 http://dx.doi.org/10.1016/S2468-2667(18)30001-X

Articles

Results
Training for providers took place between April 30, 2016,
and Dec 6, 2016. 177 doctors, including nine resident
doctors training in obstetrics and gynaecology, each
participated in one of the eight Z-CAN training sessions.
Of those who completed training, 153 practising doctors
(141 obstetrician gynaecologists and 12 family doctors or
paediatricians) agreed to participate in Z-CAN, completed
proctoring visits, and received contraceptive supplies to
provide Z-CAN services. The characteristics of providers
are listed in table 1. 139 clinics across the island
participated in the Z-CAN project (figure 2). The Z-CAN
programme design, scale-up, and implementation
occurred rapidly across the island, and the first Z-CAN
contraception services were offered on May 4, 2016.
As of Aug 15, 2017, data were available for 21 124 women
who had attended an initial visit in the Z-CAN programme
(table 1). The mean age of participants was 26 years
(SD 6·66).
The distribution of contraception methods used by
women before and after joining the Z-CAN programme
is shown in figure 3. Before their initial Z-CAN visit,
most women used either no method or one of the least
effective contraceptive methods (condoms, sponge,
withdrawal, spermicide, or fertility awareness methods),
and only a small proportion of women used one of the
most effective methods (male sterilisation, intrauterine
device, or implant; figure 3). At their visit, more than
14 259 (68%) women chose and received a LARC method
and 5250 (25%) women chose oral contraceptive pills or
other moderately effective hormonal contraception (eg,
depot medroxyprogesterone acetate injection). Of the
959 (5%) women who did not receive a contraceptive
method, the most common reasons were being undecided
on method preference or not ready to receive the method
that day, pregnancy could not be ruled out, or the desired
method was not in stock (table 1). Of the 14 259 women
who chose and received a LARC method, 7167 (50%)
women received a levonorgestrel-releasing intrauterine
device, 5031 (35%) women received an etonogestrel
implant, and 2061 (14%) women received a copper
intrauterine device. Women were more likely to choose
and receive a LARC method if they had a college degree,
had no insurance, had at least one livebirth, used a most
effective contraceptive method before Z-CAN, and saw a
Z-CAN provider in private practice or a public health or
academic clinic, after adjustment for all other
characteristics (table 2). Women aged 25 years or more
and women using a moderately effective contraceptive
method before Z-CAN were less likely to choose and
receive a LARC method. Results were similar when the
analysis was restricted to women who received a
contraceptive method at their initial visit.
The satisfaction survey began on Oct 28, 2016.
By July 21, 2017, 9829 women had received invitations
to complete the patient satisfaction survey, and
3489 (36%) women had responded (2482 women

n/N (%)
Provider characteristics
Provider type
Obstetrician-gynaecologist

141/153 (92%)

Family doctor

10/153 (7%)

Paediatrician

2/153 (1%)

Practice type
Private practice
Community health centre*
Public health clinic†
Academic clinic‡

102/153 (67%)
38/153 (25%)
3/153 (2%)
10/153 (7%)

Participant characteristics
Age, years
≤20

4539/21 124 (22%)

21–24

6057/21 124 (29%)

25–34

7759/21 124 (37%)

≥35

2558/21 124 (12%)

Relationship status
Single
Married or partnered

8887/21 124 (42%)
11 979/21 124 (57%)

Education
≤12 years
College degree
Graduate degree

7895/21 124 (37%)
11 024/21 124 (52%)
1941/21 124 (9%)

Insurance status
Private or other

8813/21 124 (42%)

Public

10 786/21 124 (51%)

None

1111/21 124 (5%)

Previous livebirth
0

7762/21 124 (37%)

≥1

12 491/21 124 (59%)

Breastfeeding at time of initial visit
No
Yes

17 213/21 124 (82%)
3350/21 124 (16%)

Did not want to conceive in the next year

20 829/21 124 (95%)

Received same-day services

20 110/21 124 (95%)

Did not receive a contraceptive method at
initial visit

959/21 124 (5%)

Undecided or not ready

410/959 (43%)

Might be pregnant

217/959 (23%)

Desired method out of stock

97/959 (10%)

Medical reason

83/959 (9%)

Reason not specified

78/959 (8%)

Did not want a contraceptive method

37/959 (4%)

Continuing current method

26/959 (3%)

Pregnant

11/959 (1%)

Proportions might not add up to 100% because of missing data. *Funded by the
Health Resources and Services Administration. †Funded by the Puerto Rico
Department of Public Health. ‡Affiliated with the University of Puerto Rico.

Table 1: Characteristics of Zika Contraception Access Network (Z-CAN)
providers and the first 21 124 women enrolled in the Z-CAN programme,
as of Aug 15, 2017

www.thelancet.com/public-health Published online January 18, 2018 http://dx.doi.org/10.1016/S2468-2667(18)30001-X

Articles

Private practice clinic (n=89)
Community health centre (n=40)*
Public health clinic (n=5)
Academic clinics (n=5)

Aguadilla

Isabela

Moca
Aguada
Rincón
Añasco
Mayagüez

Quebradillas

Barceloneta

Hatillo
Arecibo

Camuy
San
Sebastián
Las
Marías
Maricao

Utuado

Lares
Adjuntas

Hormigueros
San
Peñuelas
Germán Sabana Yauco
Grande
Cabo
Guayanilla
Rojo
Lajas
Guánica

Dorado
Cataño
Loíza
Vega
Toa Baja
Manatí Baja Vega
Alta
Juan
San
Río
Florida
Toa Alta
Carolina
Morovis
Grande Luquillo
Guaynabo Trujillo
Canóvanas
Alto
Ciales
Fajardo
Corozal
Bayamón
Gurabo
Naranjito Aguas

Jayuya

Ponce

Comerío Buenas
Naguabo
Barranquitas
Juncos
Caguas
Las Humacao
Cidra
San
Villalba
Piedras
Lorenzo
Aibonito
Coamo
Cayey
Yabucoa
Juana
Patillas
Salinas
Santa
Díaz
Maunabo
Guayama Arroyo
Isabel

Culebra

Ceiba

Orocovis

Vieques

Figure 2: Puerto Rico Zika Contraception Access Network clinics
*Includes 17 community health centres and 23 satellite clinics. Source: Zika Contraception Access Network as of Sept 23, 2017.

Proportion of women (%)

Before initial Z-CAN visit
After Z-CAN visit

n=14 259
(68%)

60
n=9285
(44%)

50
40
n=5250
n=4558 (25%)
(22%)

n=6219
(29%)

20
10
0

n=601
(3%)

n=767
(4%)
Most effective

Moderately effective

Least effective

n=959
(5%)
No method

Method of contraception

Figure 3: Contraceptive method use by women before and after their initial visit to a Zika Contraception
Access Network (Z-CAN) provider in Puerto Rico, as of Aug 15, 2017 (N=21 124)
Proportions might not add up to 100% because of missing data. Most effective contraceptive methods include
intrauterine devices, implants, and partner sterilisation. Less than 1% of women using these methods will get
pregnant during the first year of typical use. Moderately effective contraceptive methods include injectables, pills,
patch, ring, and diaphragm. 6–12% of women using these methods will get pregnant during the first year of typical
use. Least effective birth control methods include male and female condoms, withdrawal, sponge, fertility
awareness methods, and spermicides. Least effective birth control methods have a failure rate of 18 or more
pregnancies per 100 women who use these methods each year. The Centres for Disease Control and Prevention
have produced an overview of the effectiveness of family planning methods. Methods provided by Z-CAN included
intrauterine devices, implants, injectables, pills, patch, ring, and male condoms.

For the effectiveness of family
planning methods see https://
www.cdc.gov/reproductivehealth/
unintendedpregnancy/pdf/
contraceptive_methods_508.pdf

responded by email invitation, 1006 women responded by
text message invitation, and one woman responded by
phone administration). We were able to link initial visit
data to survey data for 3439 (99%) respondents.

Respondents differed from non-respondents with respect
to age, insurance status, and type of method received;
compared with non-respondents, respondents overall
were slightly older, had private insurance, and chose a
more effective method during their visit. 3489 women
participated in the patient satisfaction survey, but not all
women completed every question of the survey.
3068 (93%) of the 3294 women who answered the
question about their satisfaction with services were very
satisfied, 203 (6%) women were somewhat satisfied, and
23 (1%) women were not satisfied. 3216 (93%) of the 3478
women who answered the question about receiving the
method they were most interested in after receiving
counselling did receive the method they were most
interested in. Of the 3040 women who completed every
item on the 11-item interpersonal quality of family
planning care scale, 2382 (78%) respondents rated their
care as excellent or very good on all 11 items. Results
from individual items measuring quality of care are
summarised in the appendix.

Discussion
In Puerto Rico, the combination of a high incidence of
Zika virus infection, a high incidence of unintended
pregnancy, and low use of highly effective contraception
necessitated programmatic efforts to improve contraceptive access as a primary prevention strategy to
reduce adverse pregnancy and birth outcomes related to
Zika virus infection. The Z-CAN programme shows the
feasibility of implementing a programme to increase

www.thelancet.com/public-health Published online January 18, 2018 http://dx.doi.org/10.1016/S2468-2667(18)30001-X

Articles

access to the full range of reversible contraception,
including LARC methods, within a complex public health
response. Z-CAN also shows that it is possible to build
capacity quickly with standardised and targeted training
sessions and limited mentoring of committed providers
and to provide high-quality, comprehensive contraceptive
services in an emergency response.
Contraception has an important role in the Zika
response because Zika virus infection during pregnancy
increases the risk for microcephaly and other severe birth
defects.2 Contraception could be a key response strategy
in other public health emergencies in which prenatal
exposures pose a severe risk to pregnant women and
their infants.19 Guidance for rapid reproductive health
assessment and programme implementation in
emergency settings is available, but existing tools position
contraception services as post-emergency activities rather
than services to be implemented in the emergency
phase.20 Z-CAN shows that with concerted effort,
commitment, dedicated resources, and recognition of the
benefits of giving women the option to prevent pregnancy
during a time of crisis, it is possible to prioritise and
implement effective contraceptive provision early in an
emergency response.
Contraceptive use and provision in Puerto Rico before
the Z-CAN programme was limited by policy, financial,
and logistical barriers.4,21 Most of the 21 124 women seen
by the Z-CAN programmme chose and received a LARC
method, and most of these women were not using an
effective method of contraception before Z-CAN; these
findings suggest that when barriers to access are removed
(eg, cost, limited service points, and lack of providers),
most women who wanted to prevent pregnancy during
the Zika virus outbreak chose a highly effective method
of contraception. The choice of a LARC method was
more likely in women who had previously given birth
than in nulliparous women. Intrauterine devices are
generally safe for all women, including nulliparous
women.14 Providers might have misconceptions about
the safety of intrauterine devices in nulliparous women,
which have been shown to be associated with infrequent
provision,22 emphasising the opportunity for providers to
include LARC methods in counselling and eligibility
determinations for all women seeking contraception.
Although use of LARC methods by women using
contraception in the USA is low (14%),10 our findings are
consistent with those from other demonstration
projects9,23 that removed barriers to LARC access such as
cost, provider availability, geographic access, and
comprehensive contraception counselling. Women who
chose a short-acting method were given up to 6 months
advanced supply. Women who perceived a return visit to
receive additional contraceptive supplies as a barrier
might have inadvertently been incentivised to choose a
LARC method. However, results from the patient
satisfaction survey suggested that most women left their
initial Z-CAN visit with the method they were most

LARC (n=14 259)

Other
contraceptive
method (n=6810)

Unadjusted
prevalence ratio,
95% CI

Adjusted
prevalence ratio,
95% CI*
Referent

Age, years
≤20

2930/14 125 (21%)

1594/6734 (24%)

Referent

21–24

4176/14 125 (30%)

1868/6734 (28%)

1·07, 1·03–1·10†

1·00, 0·97–1·03

25–34

5305/14 125 (38%)

2435/6734 (36%)

1·06, 1·02–1·10†

0·93, 0·90–0·97†

≥35

1714/14 125 (12%)

837/6734 (12%)

1·04, 0·98–1·10

0·85, 0·80–0·92†

Single

5717/14 106 (41%)

3148/6709 (47%)

Referent

Married or partnered

8389/14 106 (60%)

3561/6709 (53%)

1·09, 1·04–1·14†

0·99, 0·95–1·04

≤12 years

5258/14 094 (37%)

2617/6712 (39%)

Referent

College degree

7585/14 094 (54%)

3411/6712 (51%)

1·03, 1·00–1·07

1·04, 1·01–1·08†

Graduate degree

1251/14 094 (9%)

684/6712 (10%)

0·97, 0·91–1·03

1·02, 0·96–1·08

Relationship status

Education

Insurance status
Private or other

5827/13 970 (42%)

2968/6689 (44%) Referent

Referent

Public

7326/13 970 (52%)

3429/6689 (51%)

1·03, 0·97–1·09

0·97, 0·91–1·02

None

817/13 970 (6%)

292/6689 (4%)

1·11, 1·05–1·18†

1·11, 1·05–1·17†

Previous livebirth
0

4301/13 688 (31%)

3431/6511 (53%)

Referent

1 or more

9387/13 688 (69%)

3080/6511 (47%)

1·35, 1·27–1·44†

1·40, 1·31–1·48†

11 271/13 884 (81%)

5892/6626 (89%)

Referent

Yes

2613/13 884 (19%)

734/6626 (11%)

1·19, 1·14–1·24†

1·03, 0·99–1·08
Referent

Currently breastfeeding

Effectiveness of contraceptive method used before Z-CAN‡
None

6357/14 097 (45%)

2909/6683 (44%)

Referent

Least

4451/14 097 (32%)

1757/6683 (26%)

1·05, 0·98–1·11

1·05, 0·99–1·11

Moderately

2666/14 097 (19%)

1874/6683 (28%)

0·86, 0·82–0·89†

0·90, 0·86–0·94†

623/14 097 (4%)

143/6683 (2%)

1·19, 1·12–1·25†

1·13, 1·06–1·21†

Community health
clinic

2154/14 259 (15%)

1521/6810 (22%)

Referent

Private practice or
other

12 105/14 259 (85%)

5289/6810 (78%)

1·19, 1·06–1·33†

1·19, 1·07–1·33†

Most
Clinic type

Data are n/N (%) unless indicated otherwise. LARC=long-acting reversible contraceptive. *Each characteristic in the
table was adjusted for all other characteristics. †95% CI does not include 1. ‡ Least effective contraceptive methods
include condoms for men and women, withdrawal, sponge, fertility awareness methods, and spermicides. Moderately
effective contraceptive methods include injectables, pills, patch, ring, and diaphragm. Most effective contraceptive
methods include intrauterine devices, implants, and partner sterilisation. Sterilised women were not eligible for Z-CAN
services.

Table 2: Factors associated with choosing and receiving a LARC method among the first 21 124 women
enrolled in the Zika Contraception Access Network (Z-CAN) programme, as of Aug 15, 2017

interested in receiving. In the context of the Zika virus
outbreak, improved access to contraception has the
potential to decrease unintended pregnancies and the
number of adverse pregnancy and birth outcomes related
to Zika virus infection.1,4,24
On the basis of results from multiple large-scale
programmes and research studies to reduce barriers to
contraceptive access, we anticipated that Z-CAN services
would lead to an increase in LARC use. Because of their
many advantages, including high effectiveness, safety,
reversibility, user ease, high user satisfaction, and
cost-effectiveness, LARC methods are crucial in public

www.thelancet.com/public-health Published online January 18, 2018 http://dx.doi.org/10.1016/S2468-2667(18)30001-X

Articles

health efforts to decrease unintended pregnancies.
However, issues of perceived or actual provider coercion
of women to choose LARC methods (or refuse LARC
removals), particularly based on age, race, and class, have
been reported.25,26 The historical context of unethical
contraceptive practices and research in Puerto Rico and
concerns for reproductive coercion with LARC provision
were important considerations in programme design. An
important element of the Z-CAN training and proctoring
for all providers and clinic staff was to develop competency
in delivering high-quality, patient-centred contraceptive
counselling that facilitated autonomous decision
making.13 Respondents to the satisfaction survey indicated
high satisfaction with Z-CAN services, and nearly all
women received the method they were most interested in
after counselling, suggesting that participants received
high-quality and patient-centred services through Z-CAN.
The Z-CAN programme evaluation will include additional
follow-up surveys of women participating in the
programme to further assess quality of and satisfaction
with Z-CAN services.
Through partnership and collaboration with a diverse
group of stakeholders, Z-CAN reduced barriers to
contraception as part of the public health response to the
Zika virus outbreak and expanded the capacity of
Puerto Rico’s health-care system to integrate same-day
access to contraceptive services into normal clinic
practice. Z-CAN efforts to build sustainability with key
stakeholders include building the capacity of a broad
network of providers who can provide access to
contraception, raising awareness in women of
reproductive age in Puerto Rico about the availability of
contraceptive methods, expanding the number of
contraceptive service access sites, eliminating prior
authorisation requirements and cost-sharing in health
insurance plans, and discussing continued availability of
LARC methods in Puerto Rico through pricing
negotiations and development of a sustainable supply
chain with manufacturers. Although the total cost to
implement, sustain, or replicate the Z-CAN programme
is difficult to calculate, the most expensive aspects of the
programme were provision of the contraceptive methods
(almost all of which were donated in the case of Z-CAN)
and provider reimbursement for services. Different
contexts will have different cost challenges, but the
financing requirements of these crucial aspects might be
substantial and should be considered in programme
design and sustainability planning. Successful
sustainability will be achieved if the elimination of the
most pressing barriers addressed by Z-CAN is
maintained.
This programme has several strengths. To our
knowledge, Z-CAN is the first contraception access
programme developed as a primary prevention strategy
to mitigate the effect of a Zika virus outbreak, and it is the
first contraception access programme as a primary
intervention to prevent adverse pregnancy and birth
8

outcomes in the context of a public health emergency
response. The Z-CAN programme contains important
elements of both rapid programme design and
implementation and sustainability planning and can be
adapted to other settings in which improving
contraceptive access could enhance the response to an
emergency. The strong partnerships between programme
teams and stakeholders in Puerto Rico and the high
demand for contraceptive services also strengthened the
programme.
The Z-CAN programme and this study also have
several limitations. Although Z-CAN had broad coverage
across the island, the programme was not able to provide
services in municipalities without health-care
infrastructure, so some women had to travel outside
their municipality to access care. Because of the rapid
design and implementation of Z-CAN and the specific
threat of Zika virus to maternal and child health, our
results are not readily generalisable to non-emergency
situations. The response rate to the patient satisfaction
survey was low, and the results of the survey might not be
generalisable to all women who received Z-CAN services.
The programme was implemented to serve women
throughout the risk period for Zika virus transmission,
while working towards sustainability of high-quality and
accessible contraceptive services. Although the design
and implementation phases were relatively fast, ratelimiting steps (eg, design of a procurement and
distribution system for donated contraceptive methods)
slowed the delivery of services in the early phases of the
programme. In view of the challenges of procurement
and payment of LARC methods, reaching a level of
sustainability of contraceptive services that closely
mimics Z-CAN will probably be difficult.21
Z-CAN was designed as a short-term response for rapid
implementation of contraceptive services in a complex
emergency setting. Z-CAN has established an extensive
network of providers in Puerto Rico and has served more
than 21 000 women seeking to prevent pregnancy during
the risk period for Zika virus infection. The programme
might have prevented unintended pregnancies and birth
defects related to Zika virus infections during the
outbreak. Mosquito-borne transmission of Zika virus has
reached 95 countries worldwide and all but two countries
in the Latin America Caribbean Region.27 On the basis of
these preliminary results, Z-CAN is a model programme
that could be replicated or adapted in these settings as
part of emergency preparedness and response efforts.
Additionally, Z-CAN’s design and implementation could
be refined and adapted in other non-emergent settings, in
which increased access to contraception could improve
health outcomes.
Contributors
EL, LR, SH, LBZ, MTF, MIR, and ENB-B contributed to literature search,
study design, data collection, analysis, and interpretation, and manuscript
preparation. NB contributed to literature search, study design, and data
collection. MAH, JM, and DJJ contributed to literature search, study
design, data interpretation, and manuscript preparation. The Z-CAN

www.thelancet.com/public-health Published online January 18, 2018 http://dx.doi.org/10.1016/S2468-2667(18)30001-X

Articles

Working Group members collectively contributed to the literature search,
study design, intervention implementation, and data collection.
Z-CAN Working Group
Lisa Koonin, Pierina Cordero, Ricardo Torres, Brenda Rivera,
Claritsa Malave, Alicia Suarez, Yari Vale, Linette Sanchez, Brandon Talley,
Laura Angel, Reema Bhakta, Turquoise Sidibe, Zipatly V Mendoza,
Rachel Powell, Melissa Bennett, Katherine B. Simmons, Naomi Tepper,
Jamie Krashin, Anna Brittain, Euna M August, Kathryn M Curtis,
Maura Whiteman, Jackie Rosenthal, Caitlin Green, Charity Ntansah,
Anna Fulton, Heather Clayton, Esteban Galarza, Carla Agosto,
Luz Marilyn Colón López, Madelyn Rodriguez, Brian D Montalvo Martínez,
Jeamy Rodriguez Coss, Martha Cañellas Garcia, Elvin Class Rodriguez,
Juan L Cantres, Nilda Moreno Ruiz, Stephanie Rivera, Elizabeth Sotomayor,
Ricardo Melendez, Susanna N Visser, Melody Stevens, and Von Nguyen.
Declaration of interests
We declare no competing interests. The findings and conclusions in this
report are those of the authors and do not necessarily represent the
official position of the CDC.
Acknowledgments
This study was funded by the National Foundation for the Centers for
Disease Control and Prevention. Funding via the CDC Foundation was made
possible by the Bill & Melinda Gates Foundation, Bloomberg Philanthropies,
the William and Flora Hewlett Foundation, Pfizer Foundation, and the
American College of Obstetricians and Gynecologists. The CDC Foundation
also secured large-scale donations, offers of contraceptive products or
services from Allergan, Medicines360, Americares and Janssen, Bayer,
Merck & Co, Mylan, Pfizer, Teva Pharmaceuticals, Church & Dwight, RB,
The National Campaign to Prevent Teen and Unplanned Pregnancy,
Upstream USA, and MarketVision, Culture Inspired Marketing.
The Zika Contraception Access Network would not have been possible
without the support of the CDC Foundation and technical assistance from
the CDC. We acknowledge the work of (in alphabetical order) Kate Agin,
Dayna Alexander, Tanya Alvarez, Subhashini Babu, Jason Baker,
Wanda Barfield, Sarah David, Kennis Dees, Romeo Galang, Luis Garcia,
Kim Holt, Tochukwu Igbo, Jenna Klockenbrink, Kinzie Lee, Rui Li,
Bradford Lord, Karla Moreno, Elizabeth Pantino McClune, Pierce Nelson,
Verla Neslund, Kara Polen, Nicki Roth, Wendy Ruben, Samantha Sater,
Joseph Segovia, Carrie Shapiro-Mendoza, Claire Stinson, Jasmin Taylor,
Ruben Torrez, Maria del Carmen Vidal, Lee Warner, and
John Zimmerman. We acknowledge the collaborative contributions of (in
alphabetical order) the American College of Obstetricians and
Gynecologists, the Beyond the Pill Program at the Bixby Center for Global
Reproductive Health (University of California, San Francisco School of
Medicine), Health Resources and Services Administration, Health
Resources and Services Administration Office of Regional Operations, the
Puerto Rico Department of Health, Puerto Rico Obstetrics and Gynecology,
the Puerto Rico Primary Care Association, The National Campaign to
Prevent Teen and Unplanned Pregnancy, and Upstream USA.
References
1 Boulet SL, D’Angelo DV, Morrow B, et al. Contraceptive use among
nonpregnant and postpartum women at risk for unintended
pregnancy, and female high school students, in the context of Zika
preparedness—United States, 2011–2013 and 2015.
MMWR Morb Mortal Wkly Rep 2016; 65: 780–87.
2 Oussayef NL, Pillai SK, Honein MA, et al. Zika virus—10 public
health achievements in 2016 and future priorities.
MMWR Morb Mortal Wkly Rep 2016; 65: 1482–88.
3 Lozier M, Adams L, Febo MF, et al. Incidence of Zika virus disease
by age and sex—Puerto Rico, November 1, 2015–October 20, 2016.
MMWR Morb Mortal Wkly Rep 2016; 65: 1219–23.
4 Tepper NK, Goldberg HI, Bernal MI, et al. Estimating contraceptive
needs and increasing access to contraception in response to the
Zika virus disease outbreak—Puerto Rico, 2016.
MMWR Morb Mortal Wkly Rep 2016; 65: 311–14.
5 Reynolds MR, Jones AM, Petersen EE, et al. Vital signs: update on
Zika virus-associated birth defects and evaluation of all US infants
with congenital Zika virus exposure—US Zika Pregnancy Registry,
2016. MMWR Morb Mortal Wkly Rep 2017; 66: 366–73.

6
7
8
9

11
12
13
14
15
16

17
18
19
20

21
22

24
25

26
27

McNicholas C, Madden T, Secura G, Peipert JF. The contraceptive
CHOICE project round up: what we did and what we learned.
Clin Obstet Gynecol 2014; 57: 635–43.
Peipert JF, Madden T, Allsworth JE, Secura GM. Preventing
unintended pregnancies by providing no-cost contraception.
Obstet Gynecol 2012; 120: 1291–97.
Biggs MA, Rocca CH, Brindis CD, Hirsch H, Grossman D.
Did increasing use of highly effective contraception contribute to
declining abortions in Iowa? Contraception 2015; 91: 167–73.
Ricketts S, Klingler G, Schwalberg R. Game change in Colorado:
widespread use of long-acting reversible contraceptives and rapid
decline in births among young, low-income women.
Perspect Sex Reprod Health 2014; 46: 125–32.
Kavanaugh ML, Jerman J. Contraceptive method use in the
United States: trends and characteristics between 2008, 2012 and
2014. Contraception 2017; published online Oct 13.
DOI:10.1016/j.contraception.2017.10.003.
Boring CC, Rochat RW, Becerra J. Sterilization regret among
Puerto Rican women. Fertil Steril 1988; 49: 973–81.
PBS. People & Events: The Puerto Rico Pill Trials 2001. http://www.
pbs.org/wgbh//amex/pill/peopleevents/e_puertorico.html
(accessed Jan 2, 2017).
Harper CC, Rocca CH, Thompson KM, et al. Reductions in
pregnancy rates in the USA with long-acting reversible
contraception: a cluster randomised trial. Lancet 2015; 386: 562–68.
Curtis KM, Tepper NK, Jatlaoui TC, et al. U.S. medical eligibility
criteria for contraceptive use, 2016. MMWR Morb Mortal Wkly Rep
2016; 65: 1–103.
Curtis KM, Jatlaoui TC, Tepper NK, et al. U.S. selected practice
recommendations for contraceptive use, 2016.
MMWR Recomm Rep 2016; 65: 1–66.
Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG.
Research electronic data capture (REDCap)—a metadata-driven
methodology and workflow process for providing translational
research informatics support. J Biomed Inform 2009; 42: 377–81.
CDC. Public health ethics. https://www.cdc.gov/od/science/
integrity/phethics/ (accessed Jan 2, 2017).
Dehlendorf C, Henderson JT, Vittinghoff E, et al. Association of the
quality of interpersonal care during family planning counseling
with contraceptive use. Am J Obstet Gynecol 2016; 215: 78.e1–9.
Ellington SR, Kourtis AP, Curtis KM, et al. Contraceptive availability
during an emergency response in the United States.
J Womens Health (Larchmt) 2013; 22: 189–93.
UNHCR. Inter-agency field manual on reproductive health in
humanitarian settings 2010. http://www.who.int/
reproductivehealth/publications/emergencies/field_manual_rh_
humanitarian_settings.pdf (accessed Jan 2, 2017).
Gutierrez N. Understanding health care disparities in the US
territories. Arch Intern Med 2011; 171: 1579–81.
Tyler CP, Whiteman MK, Zapata LB, Curtis KM, Hillis SD,
Marchbanks PA. Health care provider attitudes and practices related
to intrauterine devices for nulliparous women. Obstet Gynecol 2012;
119: 762–71.
Secura GM, Allsworth JE, Madden T, Mullersman JL, Peipert JF.
The Contraceptive CHOICE Project: reducing barriers to
long-acting reversible contraception. Am J Obstet Gynecol 2010;
203: 115.e1–7.
Li R, Simmons KB, Bertolli J, et al. Cost-effectiveness of increasing
access to contraception during the Zika virus outbreak, Puerto Rico,
2016. Emerg Infect Dis 2017; 23: 74–82.
Higgins JA, Kramer RD, Ryder KM. Provider bias in long-acting
reversible contraception (LARC) promotion and removal:
perceptions of young adult women. Am J Public Health 2016;
106: 1932–37.
Gubrium AC, Mann ES, Borrero S, et al. Realizing reproductive
health equity needs more than long-acting reversible contraception
(LARC). Am J Public Health 2016; 106: 18–19.
CDC. CDC’s world map of areas with risk of Zika. https://wwwnc.cdc.
gov/travel/page/world-map-areas-with-zika (accessed Nov 13, 2017).

www.thelancet.com/public-health Published online January 18, 2018 http://dx.doi.org/10.1016/S2468-2667(18)30001-X

Morbidity and Mortality Weekly Report

Estimating Contraceptive Needs and Increasing Access to Contraception in
Response to the Zika Virus Disease Outbreak — Puerto Rico, 2016
Naomi K. Tepper, MD1; Howard I. Goldberg, PhD1; Manuel I. Vargas Bernal, MD2; Brenda Rivera, DVM2; Meghan T. Frey, MPH3; Claritsa Malave, MD4;
Christina M. Renquist, MPH3; Nabal Jose Bracero, MD5; Kenneth L. Dominguez, MD6; Ramon E. Sanchez, MD7; Carrie K. Shapiro-Mendoza, PhD1;
Blanca R. Cuevas Rodriguez, MS8; Regina M. Simeone, MPH3; Nicki T. Pesik, MD9; Wanda D. Barfield, MD1; Jean Y. Ko, PhD1; Romeo R. Galang, MD6,10;
Janice Perez-Padilla, MPH11; Kara N.D. Polen, MPH3; Margaret A. Honein, PhD3; Sonja A. Rasmussen, MD12; Denise J. Jamieson, MD1

On March 25, 2016, this report was posted as an MMWR
Early Release on the MMWR website (http://www.cdc.gov/mmwr).
Zika virus is a flavivirus transmitted primarily by Aedes species
mosquitoes. Increasing evidence links Zika virus infection
during pregnancy to adverse pregnancy and birth outcomes,
including pregnancy loss, intrauterine growth restriction, eye
defects, congenital brain abnormalities, and other fetal abnormalities (1,2). The virus has also been determined to be sexually
transmitted.* Because of the potential risks associated with Zika
virus infection during pregnancy, CDC has recommended that
health care providers discuss prevention of unintended pregnancy with women and couples who reside in areas of active
Zika virus transmission and do not want to become pregnant.†
However, limitations in access to contraception in some of
these areas might affect the ability to prevent an unintended
pregnancy. As of March 16, 2016, the highest number of Zika
virus disease cases in the United States and U.S. territories were
reported from Puerto Rico.§ The number of cases will likely rise
with increasing mosquito activity in affected areas, resulting in
increased risk for transmission to pregnant women. High rates
of unintended and adolescent pregnancies in Puerto Rico suggest that, in the context of this outbreak, access to contraception
might need to be improved (3,4). CDC estimates that 138,000
women of reproductive age (aged 15–44 years) in Puerto Rico
do not desire pregnancy and are not using one of the most
effective or moderately effective contraceptive methods,¶,**
and therefore might experience an unintended pregnancy.
CDC and other federal and local partners are seeking to expand
access to contraception for these persons. Such efforts have
the potential to increase contraceptive access and use, reduce
unintended pregnancies, and lead to fewer adverse pregnancy
and birth outcomes associated with Zika virus infection during
pregnancy. The assessment of challenges and resources related
to contraceptive access in Puerto Rico might be a useful model
for other areas with active transmission of Zika virus.

CDC, the Puerto Rico Department of Health, and partners
used a comprehensive approach, including key informant interviews and review of existing data, to gather information on contraception services in Puerto Rico, including information on
rates of unintended pregnancy, contraceptive use, contraceptive
access, and barriers to provision and use of contraception.
Discussions were conducted with federal partners, including
the Center for Medicare and Medicaid Services, the Office
of Population Affairs, and the Health Resources and Services
Administration (HRSA). Key stakeholders and family planning
providers in Puerto Rico were also consulted, including the
Puerto Rico Department of Health, the Puerto Rico Chapter
of the American College of Obstetricians and Gynecologists
(ACOG), Title X federal family planning grantees, and the
Puerto Rico Health Insurance Administration.
Because current data regarding contraceptive use prevalence
in Puerto Rico are not available, the number of women in
Puerto Rico who desire effective contraception was estimated
using several data sources. The estimated number of women
of reproductive age (15–44 years) in 2014 was obtained
from the U.S. Census Bureau.†† To determine the number
of women of reproductive age who are not using one of the
most effective or moderately effective contraceptive methods
and who might therefore have an unintended pregnancy, a
series of assumptions were made. Based on national results
from the 2013 Youth Risk Behavior Surveillance System,
50% of women aged 15–19 years were assumed to be sexually experienced, and among these, 90% were assumed not to
desire pregnancy and not to be using one of the most effective
or moderately effective contraceptive methods.§§,¶¶ Among
women aged 20–44 years, 65% were assumed to be sexually
active, not infertile, not currently pregnant, and not currently
desiring to become pregnant (5). The number of women aged
20–44 years who might have an unintended pregnancy was
estimated by assuming that 65% were not sterilized (6), and

* http://www.cdc.gov/mmwr/volumes/65/wr/mm6508e2.htm.
† http://www.cdc.gov/mmwr/volumes/65/wr/mm6505e2.htm.
§ http://www.cdc.gov/zika/geo/united-states.html.
¶ http://www.cdc.gov/reproductivehealth/unintendedpregnancy/pdf/
contraceptive_methods_508.pdf.
** Most effective = sterilization, intrauterine device, contraceptive implant;
moderately effective = injectable contraceptive, oral contraceptive,
contraceptive patch, or contraceptive vaginal ring.

†† http://www.census.gov.
§§
http://www.cdc.gov/mmwr/preview/mmwrhtml/ss6304a1.htm.
¶¶
Estimated number of sexually active women aged 15–19 years who might
have an unintended pregnancy = (no. women aged 15–19 years) x (50%
sexually active) x (90% not desiring pregnancy, not infertile, not using
effective contraception).

US Department of Health and Human Services/Centers for Disease Control and Prevention

MMWR / April 1, 2016 / Vol. 65 / No. 12

311

Morbidity and Mortality Weekly Report

that among those, 33% are not using one of the most effective
or moderately effective reversible contraceptive methods (5).***
To estimate the percentage distribution of desired contraceptive methods that might be needed in Puerto Rico, data from
the Contraceptive CHOICE project, which was designed to
remove the financial barriers to contraception, offer all methods
and emphasize the most effective methods of birth control,
and reduce unintended pregnancy in the St. Louis, Missouri
area during 2007–2011,††† was used. In this project, women
desiring reversible contraception were offered any Food and
Drug Administration–approved contraceptive method at no
cost along with counseling to promote the use of long-acting
reversible contraceptive (LARC) methods (intrauterine devices
[IUDs] and hormonal contraceptive implants), because these
are the most effective reversible methods. Seventy-five percent
of the general study population and 72% of adolescents aged
15–19 years chose a LARC method, resulting in decreases in
adolescent and unintended pregnancy (7,8). Demonstration
projects in Iowa and Colorado, also designed to increase use
of LARC methods, have similarly resulted in increased use
of LARCs and decreases in unintended pregnancy.§§§,¶¶¶
Assuming a distribution of desired methods similar to that
observed in the CHOICE project (7,8), if barriers to access
were removed, the total number of contraceptive products
needed in Puerto Rico to supply all women of reproductive
age who are currently not using one of the most effective or
moderately effective contraceptive methods and who do not
want to become pregnant was estimated.
Approximately 715,000 women aged 15–44 years reside in
Puerto Rico, and there were approximately 34,000 births in
2014 (3). A 2008 hospital-based survey of postpartum women
in Puerto Rico indicated that 65.5% of pregnancies were unintended in Puerto Rico, compared with 51% in a probability
sample of the general U.S. population (the 50 U.S. states and
the District of Columbia), according to the 2008 National
Survey of Family Growth (4,9). In 2014, among women aged
15–19 years, the birth rate was almost twice as high (40/1,000)
in Puerto Rico as in the U.S. overall (24/1,000) (3).
The most recent population-based estimates of contraceptive use in Puerto Rico, from a 2002 Behavioral Risk Factor
Surveillance System survey, found that among women aged
18–44 years who used contraception, tubal ligation was the
most frequently reported method, used by 46% of women,
*** Estimated number of sexually active women aged 20–44 years who might
have an unintended pregnancy = (no. women aged 20–44 years) x (65%
sexually active, not infertile, not currently pregnant, not desiring pregnancy)
x (65% not sterilized) x (33% not using effective reversible contraception).
††† http://www.choiceproject.wustl.edu.
§§§ http://www.astho.org/Maternal-and-Child-Health/Long-Acting-ReversibleContraception/Iowa-Initiative-Title-X-Issue-Brief/.
¶¶¶ https://www.colorado.gov/pacific/cdphe/reducing-unintended-pregnancy.

312

MMWR / April 1, 2016 / Vol. 65 / No. 12

followed by oral contraceptives (19%), condoms (11%),
calendar-based contraceptive methods (10%), vasectomy
(6%), depot medroxyprogesterone acetate (DMPA) (3%),
and IUDs (1%) (6). More recent information on services
provided by La Asociación Puertorriqueña Pro Bienestar de
la Familia (PROFAMILIA), a private non-profit organization that provides reproductive health care to a largely low
income population in Puerto Rico, indicated that among
approximately 44,000 women receiving contraceptive care
in 2009, 80% received oral contraceptives, 8% received the
transdermal contraceptive patch, 6% received condoms, 3%
received DMPA, and <1% received an IUD (4).
Women access contraception at various sites in Puerto
Rico, including community health clinics, private medical
offices, university clinics, and Title X family planning clinics
(Manuel Vargas, MD, MPH, Puerto Rico Department of
Health; Claritsa Malave, MD, MPH, HRSA; personal communications, 2016). Despite the availability of these resources,
barriers exist to providing optimal contraceptive coverage. Key
stakeholders in Puerto Rico identified the need for increased
contraceptive supplies, family planning delivery sites, training
for providers on LARC insertion, education for women and
men on effective contraception to reduce unintended pregnancy, and decreased financial and administrative barriers for
providers and patients (Manual Vargas, MD, MPH, Puerto
Rico Department of Health; Claritsa Malave, MD, MPH,
HRSA; Nabal Bracero, MD, ACOG Puerto Rico Section;
Ramon Sanchez, MD, MPH, Clinica Preven; Blanca Cuevas,
MS, PROFAMILIA; personal communications, 2016).
Coverage for all contraceptive methods by federal and private
insurers is not universal in Puerto Rico. Certain contraceptive
methods can be unaffordable for providers and patients, which
has resulted in limited availability of more effective contraceptive options such as LARCs that have higher up-front costs
(Manuel Vargas, MD, MPH, Puerto Rico Department of
Health; personal communication, 2016). In addition, the cost
of IUD and hormonal implant insertion might not be fully
covered by public or private insurance, which might also deter
women from seeking LARCs. Because of cost, these methods
are often not available in physician offices or pharmacies, and
therefore most women receive oral contraceptives, DMPA,
or condoms. A lack of availability in hospitals has also led
to missed opportunities for postpartum initiation of LARCs
(Nabal Bracero, MD, MPH, ACOG Puerto Rico Section;
personal communication, 2016). The number of health care
providers who offer contraception, specifically IUDs and
contraceptive implants, has been limited by lack of training
and reimbursement (Nabal Bracero, MD, MPH, ACOG
Puerto Rico Section; Manuel Vargas, MD, MPH, Puerto Rico
Department of Health; personal communications, 2016).

US Department of Health and Human Services/Centers for Disease Control and Prevention

Morbidity and Mortality Weekly Report

Women typically do not choose LARC methods because of
this lack of availability, as well as a general lack of knowledge
about these methods (Ramon Sanchez, MD, MPH, Clinica
Preven; personal communication, 2016).
Among the 715,000 women of reproductive age in
Puerto Rico, an estimated total of 138,000, or nearly 1 in
5 women, including 55,000 aged 15–19 years and 83,000
aged 20–44 years, do not want to become pregnant, are not
using one of the most effective or moderately effective contraceptive methods, and could therefore have an unintended
pregnancy. Applying the distribution of methods observed in
the CHOICE project, there is an estimated unmet need for
IUDs for 68,000 women, hormonal contraceptive implants
for 33,000 women, DMPA for 11,000 women, oral contraceptives for 14,000 women, vaginal rings for 9,000 women,
and contraceptive patches for 3,000 women (Table). The
estimated needs for a year are 68,000 IUDs, 33,000 hormonal
contraceptive implants, 44,000 DMPA doses, 168,000 oral
contraceptive pill packs, 108,000 vaginal rings, and 36,000
contraceptive patches.
Discussion

Reducing the rate of unintended pregnancy is a public
health priority because unintended pregnancies can be associated with delayed entry into prenatal care, decreased smoking
cessation, and increased incidence of low birthweight (10),
with attendant negative health consequences for mother and
infant. Prevention of unintended pregnancies in the context
of a Zika virus outbreak is especially important to reducing
the likelihood of congenital infections. Removing barriers to
contraception, such as cost, access, and lack of knowledge,
can lead to increased use of the most effective contraceptive
methods and reduced rates of unintended pregnancy, which
would result in fewer adverse pregnancy and birth outcomes
associated with Zika virus disease during pregnancy.

CDC and other partners have initiated multiple approaches
to address some of these barriers. Current information on
contraceptive use and unmet need is important, and efforts
are underway to conduct reproductive health surveys in Puerto
Rico to obtain this information. Approaches to increasing
access to effective contraceptive methods at no or reduced cost
are being explored. Education of providers is being conducted
through outreach sessions designed to disseminate information
about prevention of adverse outcomes associated with Zika
virus infection during pregnancy. Training of providers on
insertion of IUDs and contraceptive implants can be implemented using resources from professional organizations such as
ACOG and the University of Puerto Rico. Ongoing education
about effective use of contraception can be enhanced through
health care providers, counselors in community health centers,
home visiting nurses, and schools.
The findings in this report are subject to at least four limitations. First, no recent information was available regarding
the proportion of women of reproductive age in Puerto Rico
using specific contraceptive methods. Therefore, estimates of
contraceptive need were derived from 2002 data, highlighting the urgent need for reproductive health surveys in Puerto
Rico and other Zika-affected areas to better estimate unmet
contraceptive need. Second, contraceptive preferences were
extrapolated from the CHOICE project, and might not represent preferences in Puerto Rico or other populations, because of
demographic and cultural differences. However, demonstration
projects from other populations in the United States have similarly demonstrated high preference for LARC methods when
common barriers, including cost, availability, and knowledge,
were removed. Third, pregnancy intentions might change as a
result of the Zika virus outbreak; therefore assumptions about
pregnancy desires might not be accurate. Finally, most of the
information on contraceptive access and barriers was obtained
by nonsystematic personal communications with key leaders
and stakeholders.

TABLE. Estimated contraception needs required to supply all women who desire to avoid pregnancy,* by contraceptive method — Puerto Rico, 2016
Age group (yrs)
Percent
distribution†

Approximate no.
of women

Percent
distribution§

Approximate no.
of women

Total no. of
women

Total no. of
contraceptives
needed for
1 yr supply

37
35
9
12
5
2
100

20,000
19,000
5,000
7,000
3,000
1,000
55,000

58
17
7
9
7
2
100

48,000
14,000
6,000
7,000
6,000
2,000
83,000

68,000
33,000
11,000
14,000
9,000
3,000
138,000

68,000
33,000
44,000
168,000
108,000
36,000
457,000

15–19
Contraceptive method
Intrauterine devices
Contraceptive implants
Depot medroxyprogesterone acetate
Oral contraceptives
Contraceptive vaginal ring
Contraceptive patch
Total

20–44

* Includes women who are sexually active, fertile, and not sterilized nor using one of the most effective or moderately effective reversible contraceptive methods.
† Percent of contraceptive methods = distribution observed in CHOICE project for women aged 15–19 years (http://www.nejm.org/doi/pdf/10.1056/NEJMoa1400506).
§ Percent of contraceptive methods = distribution observed in CHOICE project for women aged 20–44 years (http://europepmc.org/articles/pmc4216614).

US Department of Health and Human Services/Centers for Disease Control and Prevention

MMWR / April 1, 2016 / Vol. 65 / No. 12

313

Morbidity and Mortality Weekly Report

Acknowledgments

Summary
What is already known about this topic?

Susan B. Moskosky, MS, Office of Population Affairs; Loretta
Gavin, PhD, Office of Population Affairs; Michael J. Melendez,
Centers for Medicare and Medicaid Services; Ivelisse M. Salce,
Centers for Medicare and Medicaid Services; Michele Lawler, Health
Resources and Services Administration.

Zika virus infection during pregnancy has been linked to
adverse pregnancy and birth outcomes, including pregnancy
loss, intrauterine growth restriction, and congenital brain
abnormalities. As of March 2016, Puerto Rico had the highest
number of cases of Zika virus disease in the United States and
its territories. Women residing in areas with active Zika virus
transmission who do not desire pregnancy need access to
effective and affordable contraception.

1Division

What is added by this report?
Approximately two thirds of pregnancies in Puerto Rico are
unintended. An estimated 138,000 women of reproductive
age (15–44 years) in Puerto Rico do not desire pregnancy and
are not using an effective contraceptive method. Access to
contraception is constrained by limited availability, especially of
highly effective long-acting reversible contraceptives, high cost,
incomplete insurance coverage, and lack of trained providers.
To adequately prevent unintended pregnancies, there is an
estimated need for IUDs for 68,000 women, contraceptive
implants for 33,000 women, depot medroxyprogesterone
acetate for 11,000 women, oral contraceptives for 14,000
women, vaginal rings for 9,000 women, and contraceptive
patches for 3,000 women.
What are the implications for public health practice?
Removing barriers to contraception, such as cost, limited
access, and lack of knowledge, could lead to increased use of
highly effective contraceptive methods and reduced rates of
unintended pregnancy, resulting in fewer adverse pregnancy
and birth outcomes in the context of a Zika virus disease
outbreak. This assessment of the resources and challenges in
Puerto Rico related to contraceptive access might be a useful
model for other areas with active transmission of Zika virus.

A collaborative and coordinated response is required from
federal and local partners as well as other stakeholders, such
as academic and professional organizations, private insurance
companies, schools, and community leaders, to ensure access to
contraception for women who desire to avoid pregnancy during the Zika outbreak in Puerto Rico and other affected areas.
Increasing reimbursement and reducing costs for contraceptive
services would support access. Efforts to increase opportunities
for health care provider training on LARC insertion are needed.
Education opportunities should be increased through health
care providers, health educators, community leaders, schools,
and other outreach mechanisms. This assessment of resources
and challenges related to contraceptive access performed for
Puerto Rico might be a useful model for other areas with active
transmission of Zika virus.

314

MMWR / April 1, 2016 / Vol. 65 / No. 12

of Reproductive Health, National Center for Chronic Disease
Prevention and Health Promotion, CDC; 2Puerto Rico Department of Health;
3Division of Congenital and Developmental Disorders, National Center on
Birth Defects and Developmental Disabilities, CDC; 4Health Resources and
Services Administration, Office of Regional Operations, Region II, Puerto Rico;
5University of Puerto Rico and Puerto Rico Section of the American College of
Obstetricians and Gynecologists; 6Division of HIV/AIDS Prevention, National
Center for HIV/AIDS, Viral Hepatitis, STD, and Tuberculosis Prevention,
CDC; 7University of Puerto Rico Family Planning Program Title X-Clinica
Preven; 8PROFAMILIAS, Puerto Rico; 9Office of the Director, National Center
for Emerging and Zoonotic Infectious Diseases, CDC; 10Epidemic Intelligence
Service, CDC; 11Division of Vector-Borne Diseases, National Center for
Emerging and Zoonotic Infectious Diseases, CDC; 12Division of Public
Health Information Dissemination, Center for Surveillance, Epidemiology,
and Laboratory Services, CDC.
Corresponding author: Naomi K. Tepper, MD, [email protected],
770-488-7100.

References
1. Brasil P, Pereira JP Jr, Raja Gabaglia C, et al. Zika virus infection in
pregnant women in Rio de Janeiro—preliminary report. N Engl J Med
2016;NEJMoa1602412. Published online March 4, 2016. http://dx.doi.
org/10.1056/NEJMoa1602412
2. Meaney-Delman D, Rasmussen SA, Staples JE, et al. Zika virus and
pregnancy: what obstetric health care providers need to know. Obstet
Gynecol 2016. Published online February 17, 2016.
3. Hamilton BE, Martin JA, Osterman MJ, Curtin SC, Matthews TJ.
Births: final data for 2014. Natl Vital Stat Rep 2015;64:1–64. http://
www.cdc.gov/nchs/data/nvsr/nvsr64/nvsr64_12.pdf
4. Puerto Rico Title V Application, 2010–2011. https://mchdata.
hrsa.gov/TVISReports/Documents/NeedsAssessments/2011/
PR-NeedsAssessment.pdf
5. Daniels K, Daugherty J, Jones J, Mosher W. Current contraceptive use
and variation by selected characteristics among women aged 15–44:
United States, 2011–2013. Natl Health Stat Report 2015. http://www.
cdc.gov/nchs/data/nhsr/nhsr086.pdf
6. Bensyl DM, Iuliano DA, Carter M, Santelli J, Gilbert BC. Contraceptive
use—United States and territories, Behavioral Risk Factor Surveillance
System, 2002. MMWR Surveill Summ 2005;54(No.SS-6).
7. Peipert JF, Madden T, Allsworth JE, Secura GM. Preventing unintended
pregnancies by providing no-cost contraception. Obstet Gynecol
2012;120:1291–7.
8. Secura GM, Madden T, McNicholas C, et al. Provision of no-cost,
long-acting contraception and teenage pregnancy. N Engl J Med
2014;371:1316–23. http://dx.doi.org/10.1056/NEJMoa1400506
9. Finer LB, Zolna MR. Shifts in intended and unintended pregnancies in the
United States, 2001-2008. Am J Public Health 2014;104(Suppl 1):S43–8.
http://dx.doi.org/10.2105/AJPH.2013.301416
10. Gipson JD, Koenig MA, Hindin MJ. The effects of unintended
pregnancy on infant, child, and parental health: a review of
the literature. Stud Fam Plann 2008;39:18–38. http://dx.doi.
org/10.1111/j.1728-4465.2008.00148.x

US Department of Health and Human Services/Centers for Disease Control and Prevention

Cost-effectiveness of Increasing
Access to Contraception during
the Zika Virus Outbreak,
Puerto Rico, 2016
Rui Li, Katharine B. Simmons, Jeanne Bertolli, Brenda Rivera-Garcia, Shanna Cox,
Lisa Romero, Lisa M. Koonin, Miguel Valencia-Prado, Nabal Bracero, Denise J. Jamieson,
Wanda Barfield, Cynthia A. Moore, Cara T. Mai, Lauren C. Korhonen, Meghan T. Frey,
Janice Perez-Padilla, Ricardo Torres-Muñoz, Scott D. Grosse

We modeled the potential cost-effectiveness of increasing
access to contraception in Puerto Rico during a Zika virus
outbreak. The intervention is projected to cost an additional
$33.5 million in family planning services and is likely to be
cost-saving for the healthcare system overall. It could reduce Zika virus–related costs by $65.2 million ($2.8 million
from less Zika virus testing and monitoring and $62.3 million
from avoided costs of Zika virus–associated microcephaly
[ZAM]). The estimates are influenced by the contraception
methods used, the frequency of ZAM, and the lifetime incremental cost of ZAM. Accounting for unwanted pregnancies
that are prevented, irrespective of Zika virus infection, an
additional $40.4 million in medical costs would be avoided
through the intervention. Increasing contraceptive access
for women who want to delay or avoid pregnancy in Puerto
Rico during a Zika virus outbreak can substantially reduce
the number of cases of ZAM and healthcare costs.

ika virus infection during pregnancy can cause microcephaly with severe brain damage in the fetus
(referred to here as Zika virus–associated microcephaly
[ZAM]) and is linked to pregnancy loss and to problems
in infants, including eye defects, hearing loss, and impaired growth (1). Zika virus is a flavivirus transmitted
primarily by infected Aedes species mosquitos (2). Zika
virus can also be sexually transmitted (3). Puerto Rico
Author affiliations: Centers for Disease Control and Prevention,
Atlanta, Georgia, USA (R. Li, K.B. Simmons, J. Bertolli,
S. Cox, L. Romero, L.M. Koonin, D.J. Jamieson, W. Barfield,
C.A. Moore, C.T. Mai, L.C. Korhonen, M.T. Frey, J. Perez-Padilla,
S.D. Grosse); Puerto Rico Department of Health, San Juan,
Puerto Rico (B. Rivera-Garcia, M. Valenica-Prado,
R. Torres-Muñoz); University of Puerto Rico and Puerto Rico
Section of the American College of Obstetricians and
Gynecologists, San Juan (N. Bracero)
DOI: http://dx.doi.org/10.3201/eid2301.161322
74

has the largest number of Zika virus disease cases in the
United States and its territories (4) and, based on extrapolations from the experiences of other countries with Zika
virus outbreaks, will probably experience large numbers
of Zika virus–exposed pregnancies (5).
A primary strategy to reduce Zika virus–associated adverse pregnancy outcomes is to assist women who want to
delay or avoid pregnancy. An estimated 65% of pregnancies in Puerto Rico are unintended (unwanted or mistimed),
compared with 45% in the continental United States (2,6).
Women in Puerto Rico face multiple barriers to contraceptive use, including high out-of-pocket costs, a shortage of
contraceptive supplies, lack of education about options,
and a limited number of family planning delivery sites (2).
In response to the Zika virus outbreak, the Centers for
Disease Control and Prevention and other federal and local
partners are seeking to improve access to contraception for
women in Puerto Rico who desire it but encounter barriers to accessing the full range of contraception methods,
including long-acting reversible contraceptives (LARCs).
The objective of this analysis was to estimate the potential
cost-effectiveness of increasing access to contraception in
Puerto Rico during the 2016 Zika virus outbreak.
Methods
We constructed a decision tree cost-effectiveness model for a target population of 163,000 women who at the
time of the intervention are sexually active with a male
partner, fertile, not desiring pregnancy within the next 12
months, and not using permanent contraception methods
(e.g., tubal ligation and vasectomy) (online Technical Appendix Table and Figure 1, http://wwwnc.cdc.gov/EID/
article/23/1/16-1322-Techapp1.pdf). In the no intervention scenario, no changes in contraceptive use distributions
from the status quo are expected to occur. In the intervention scenario, women in Puerto Rico are assumed to

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 23, No. 1, January 2017

Contraception during Zika Outbreak, Puerto Rico

have same-day access to contraception methods, including
LARC, with no out-of-pocket costs. In addition, healthcare providers would be trained to provide client-centered
contraceptive counseling and outreach so that women have
the information they need to make an informed choice on
the contraception method that is best for them. The model
specifies contraceptive method use distribution, unintended pregnancy events, and the frequency of ZAM (online
Technical Appendix Figure 1).
We assumed an intervention in place throughout a
year-long Zika virus outbreak in Puerto Rico. We evaluated
the costs and outcomes of increased access to contraception compared with no intervention (i.e., status quo). Output measures included numbers of ZAM cases prevented,
including stillbirths, elective terminations, and live-born
infants, and healthy life years (HLY) gained. Economic
benefits of the intervention included avoided costs from
ZAM cases prevented and costs avoided for monitoring
for Zika virus–exposed pregnancies and infants born from
Zika virus–infected mothers. In addition, the avoided cost
of prenatal, delivery, postpartum, and neonatal care associated with avoided unwanted pregnancies was considered
an economic benefit. In cost-effectiveness analyses, if total
avoided cost exceeds the cost of an intervention that improves health, the intervention is considered cost-saving.
For scenarios with positive net costs, we reported the incremental cost-effectiveness ratio (ICER), which is the net
cost per HLY gained in comparison to the status quo.
Independent of Zika virus–exposed pregnancies and
ZAM, unintended pregnancy is associated with adverse
maternal and child health outcomes. Because roughly
60% of unintended pregnancies are classified as mistimed,
which might result in a delayed rather than avoided pregnancy, with the same costs occurring later (7), we only estimated avoided medical costs from prevention of the 40%
of unintended pregnancies presumed to be not desired at
a later time irrespective of Zika virus infection. However,
we included all ZAM cases prevented during the intervention period.
Contraception Use with and without the Intervention

We estimated the inputs for the decision-tree model
and their sources (Table 1, http://wwwnc.cdc.gov/EID/
article/23/1/16-1322-T1.htm). In the no intervention scenario, we took the distribution of women in the target population by use of different types of reversible contraceptives
(or no use) from a 2002 survey administered in Puerto Rico
and adjusted it to reflect the 36% decrease in fertility rates
in Puerto Rico during 2002–2015 (8,23,24).
For the main intervention scenario, we assumed that
50% of no contraception users, 60% of less-effective contraceptive method users, and 100% of moderately effective contraceptive method users would visit a healthcare

provider during the intervention period and be counseled
about contraception use (Table 1). The first 2 percentages
are roughly twice the percentages of women reported in the
2011–2013 US National Survey on Family Growth to have
received contraceptive services (contraception or counseling) within the past year because we assumed that, during
the Zika virus outbreak, more women and providers would
discuss contraception; virtually all moderately effective
method users were assumed to see providers to obtain contraceptive prescriptions.
For the main scenario, we also assumed, optimistically, that 50% of women in the target population who receive contraceptive services during the Zika virus outbreak
would be willing to change to a more effective contraceptive method, evenly divided between moderately effective
and highly effective methods. We applied data from the
Contraceptive CHOICE Project (67% of participants used
LARC and 33% used moderately effective methods) (9) to
the 40% of women assumed to not want to be pregnant;
we assumed 20% of other women not intending pregnancy
would use LARC. We further assumed that 30% of moderately effective contraception users would also choose to
use condoms (dual-method use) under the intervention,
based on a study reporting dual-method use among persons
at risk for HIV (25).
Epidemiologic Model Input Parameters

We calculated method-specific annual pregnancy rates by
applying failure rates of contraception methods under typical use (10), in combination with information on estimated
numbers of unintended pregnancies, to adjust for other factors influencing pregnancy risk (19). We estimated the proportion of fetal losses among unintended pregnancies from
data for the Caribbean region, including Puerto Rico (12),
and calculated the proportion of induced abortion among
unintended pregnancies from a survey conducted in Puerto
Rico in 2001 (the latest year for which data were available)
(11). We assumed that the distribution of fetal loss and induced abortions in unintended pregnancies unaffected by
ZAM would not be altered by the Zika virus outbreak or
the intervention.
For adverse pregnancy and birth outcomes associated
with Zika virus, we only considered ZAM and associated
brain anomalies, including live births, stillbirths, and terminations attributable to prenatal diagnosis. Although Zika
virus can cause brain lesions and dysfunction in fetuses and
newborns who do not have microcephaly (26), we lacked
the data to model their prevalence and cost. In the main
analysis, we assumed 58 cases of ZAM per 10,000 live
births (range 32–86/10,000) based on a modeling study
that considered data from other mosquitoborne illnesses
in Puerto Rico and Zika virus outbreaks in other locations
(5). We assumed a pregnancy loss rate of 35% among Zika

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 23, No. 1, January 2017

RESEARCH

virus–exposed fetuses with diagnosed birth defects based
on cases in the US Zika Pregnancy Registry as of July 21,
2016 (14).
A summary measure of population health impact is
healthy life expectancy at birth. We projected gains in
HLY by multiplying total cases of ZAM prevented by 30.0,
which is the average number of quality-adjusted life-years
at birth in the United States for an infant without severe microcephaly (15) and the estimated loss in disability-adjusted life years from microcephaly (27). We multiplied 30.0
by the sum of live births and fetal losses associated with
ZAM to calculate gains in HLY. We included fetal losses
in the HLY calculations because in the absence of ZAM
those pregnancies would have resulted in live births, with
the same healthy life expectancy as other children (15).
Cost Parameters

We conducted the analysis from a healthcare system perspective that includes direct medically related costs regardless of payer. We used payments from private insurance
because payments from Medicaid might underestimate the
cost of healthcare (28). Intervention costs included program
costs of training providers, patient educational materials,
outreach/media campaigns on the availability of contraceptives services, and program coordination and the incremental costs of family planning services. The latter comprised
the costs of contraception methods and related office visits
and services (e.g., insertion and removal of LARC for new
method users resulting from the intervention and the cost
of more intensive counseling for all women receiving contraceptive services during the intervention). We took the
1-year costs for contraception methods from the literature
(16,29) and based the other program costs on the estimated
costs for a pilot program planned to increase access to contraception in Puerto Rico as part of the current Zika virus
outbreak response (30). We did not apply a discount rate
to intervention costs because of the time horizon of 12
months.
Zika virus–related costs prevented by this intervention
were in 2 parts: 1) costs for Zika virus testing and monitoring for Zika virus–exposed pregnancies and infants, and
2) costs of ZAM cases (Table 1). The cost estimates for
testing and monitoring presumed 100% adherence by clinicians and patients to recommendations (20–22).
The lifetime cost per live-born infant with ZAM includes direct medical and nonmedical costs. ZAM is among
the most severe types of microcephaly and is associated
with loss of brain tissue volume, increased fluid spaces,
and intracranial calcifications. All 3 cases of live-born infants with ZAM in French Polynesia demonstrated severe
neurologic outcomes with delayed cognitive development
(26). On the basis of expert opinion, infants with ZAM
who survive the neonatal period would be expected to have
76

neurologic dysfunction consistent with severe cerebral palsy within 1–2 years of birth.
As a proxy for the medical cost of ZAM, we used the
estimated cost of treating infants with microcephaly associated with a diagnosis of symptomatic congenital cytomegalovirus (CMV). We used the MarketScan Commercial Database (Truven Health Analytics) with a sample of
≈100 million US residents covered by employer-sponsored
insurance at any time during 2009–2014. We used average
costs for 4 newborn infants with diagnoses of microcephaly and CMV who survived and were enrolled in a health
plan for >3 years. For the direct nonmedical cost of ZAM,
we used the estimated cost for supportive care for children
with severe congenital brain injury, both paid care and unpaid care. The total lifetime cost for surviving infants with
ZAM was estimated at $3.8 million per infant, taking into
account infant and child mortality and discounting of costs
in future years at a 3% rate per year; the sum of undiscounted costs for children who survive to adulthood might
reach $10 million.
We determined the estimated non–Zika virus–related
medical costs associated with women’s prenatal care, labor
and delivery, and postpartum care for pregnancies ending
in live birth and neonatal care from a study of US commercial health plan expenditures (17). Estimates for costs associated with pregnancies ending in induced abortion were
based on our analyses of commercial claims data (Table 1).
Sensitivity Analyses

Because many parameters used in the model are uncertain,
we conducted sensitivity analyses on selected parameters,
including different scenarios for the baseline and postintervention contraception use distributions in Puerto Rico. We
tested alternate baseline contraception use distributions in
Puerto Rico for women at risk for unintended pregnancy by
using the actual distribution of method use reported in 2002
(8) and among women attending Title X clinics in Puerto
Rico in 2014 (31). For the postintervention contraception
use distribution, we tested scenarios assuming different
proportions of women receiving contraceptive services
from a healthcare provider, different levels of willingness
to switch to a more effective method, and different shares of
moderately effective and highly effective methods among
switchers. Other parameters evaluated during sensitivity
analysis included the incidence of ZAM during the Zika
virus outbreak in Puerto Rico, percentage of pregnancies
with ZAM terminated, the cost of caring for a live-born
infant with microcephaly, and the cost of the intervention.
We conducted sensitivity analyses in which we altered
selected assumptions. In one, we annualized the cost of
LARC devices considering the expected duration of method use. In another, we adjusted observed data on US healthcare and supportive care costs to the generally lower levels

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 23, No. 1, January 2017

Contraception during Zika Outbreak, Puerto Rico

of prices in Puerto Rico market by applying conversion factors of ratios of healthcare spending per capita and wages of
nurse assistants between the United States and Puerto Rico
(32,33). We also conducted a probabilistic sensitivity analysis by using Monte Carlo simulation (10,000 draws) that
assumed different distributions for all the parameters used
in the model (Table 1). All analyses were conducted using
TreeAge Pro 2016 software (TreeAge Software, Williamstown, MA, USA) and Excel 2013 (Microsoft, Redmond,
WA, USA). All costs were adjusted to 2014 US dollars by
using the health component of the Personal Consumption
Expenditures price index (34).
Results
In the main scenario, we predict the intervention would
prevent 25 cases of ZAM among unintended pregnancies
avoided, of which 16 would have resulted in live births
(Table 2). The incremental intervention cost of US $33.5
million (i.e., $206 per member of target population) relative to no intervention (status quo) is more than offset by
$65.2 million in avoided Zika virus–associated costs, $2.8
million from extra testing and monitoring for pregnant

women and infants for Zika virus–exposed pregnancies
avoided, and $62.3 million from ZAM cases prevented.
The net savings from Zika virus–associated costs alone is
$31.7 million.
The number of ZAM cases prevented and Zika virus–
associated costs avoided are sensitive to the proportion of
women receiving contraceptive services and the proportion
of those women willing to switch to a more effective contraception method during the Zika virus outbreak (Figure;
Table 3). If the proportions of women receiving contraception services are assumed to be the same as estimated
for the continental United States in the National Survey of
Family Growth for 2011–2013 (i.e., 21% among no contraception users, 33% among less-effective method users,
and 97% among all moderately effective method users), 16
cases of ZAM are prevented, and the net savings is $15.4
million (Table 3). If 10% of women receiving contraceptive services switch to a more effective method, 6 cases of
ZAM are prevented, and net saving is $2.8 million. If the
intervention only shifts users of moderately effective methods to a highly effective method (no change in non-use or
use of less-effective methods), 7 ZAM cases are prevented,

Table 2. Zika virus–associated microcephaly cases and costs, as well as additional costs associated with unwanted pregnancies, with
and without intervention to increase access to contraception to women during the Zika virus outbreak, Puerto Rico, 2016, in main
scenario*†‡
Parameter
Without intervention With intervention
Difference
Prevention of ZAM and Zika virus–associated cost
Total no. ZAM cases
99
74
25
No. pregnancy terminations
28
21
7
No. stillbirths
7
5
2
No. live births
64
48
16
Cost of family planning services (under intervention also includes
$38,269,679
$71,738,133
$33,468,454
program cost)
Total Zika virus–associated cost
$256,578,162
$191,422,342
–$65,155,820
Costs of extra testing and monitoring for Zika virus during pregnancy
$11,125,061
$8,303,158
–$2,821,903
and for infants exposed in utero during Zika virus outbreak§
Direct costs of ZAM¶
$245,453,101
$183,119,184
–$62,333,917
Pregnancy terminations
$139,343
$103,956
–$35,387
Stillbirths
$40,025
$29,861
–$10,165
Live births
$245,273,733
$182,985,368
–$62,288,366
Cost savings from Zika virus–associated cost avoided only#
–$31,687,366
Prevention of unwanted pregnancies
No. of unwanted pregnancies**
11,995
8,949
3,046
No. induced abortions
3,385
2,525
860
No. spontaneous abortions and fetal deaths
1,679
1,253
426
No. unwanted live births
6,856
5,117
1,739
Medical cost for unwanted pregnancy
$159,074,573
$118,722,504
–$40,352,069
Net cost savings from avoiding both Zika virus–associated cost and
–$72,039,435
unwanted pregnancy cost††
*ZAM, Zika virus–associated microcephaly.
†The numbers in the columns and rows might not exactly match because of rounding.
‡Target population size: 163,000 women who do not intend to become pregnant during Zika virus outbreak. Women of reproductive age in Puerto Rico
who are sexually active with a male partner, fertile, not desiring pregnancy, and not using permanent contraception methods (e.g., tubal ligation and
vasectomy).
§Only including cost of testing for Zika virus and monitoring for exposed infants without ZAM; testing costs for infants with ZAM are included in the direct
costs of ZAM.
¶From healthcare system perspective, includes direct medical and medical-related costs, including supportive care for persons with ZAM, even if the cost
might not be paid by healthcare payers or delivered by healthcare providers.
#Total Zika virus–associated cost avoided (absolute value) minus the additional cost of family planning service under intervention compared with no
intervention.
**Unwanted pregnancies which are not desired in the future (assuming 60% of unintended pregnancies are mistimed), irrespective of Zika virus infection
††Absolute value of net medical cost for unwanted pregnancy plus absolute value of net cost savings from Zika virus–associated costs avoided.

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 23, No. 1, January 2017

RESEARCH

Figure. Sensitivity analysis indicating
the effect of changes of assumptions on
the number of ZAM cases prevented in a
proposed intervention to increase access
to contraception to women during the Zika
virus outbreak, Puerto Rico, 2016. LARC,
long-acting reversible contraceptive; ZAM,
Zika virus–associated microcephaly.

with an ICER of $24,608/HLY gained. Increasing the proportion of dual-method users increases the number of cases
of ZAM prevented and net savings attributable to higher
contraception effectiveness. The results are also sensitive
to the prevalence of ZAM among mid-trimester pregnancies, the percentage of ZAM cases resulting in live-born infants, lifetime cost per live-born infant with ZAM, and the
intervention cost. If we adjust US cost estimates for lower
prices in Puerto Rico while keeping intervention costs at
US prices, net savings are $1.7 million. In all but 1 of the
scenarios tested, the intervention is cost-saving.
A probabilistic sensitivity analysis scatter graph shows
that most of the model simulations result in ICERs in the
lower right quadrant with lower costs and better health outcomes (online Technical Appendix Figure 2). Specifically,
the intervention is cost-saving in 92.11% of the 10,000 iterations, and in 98.10% of the iterations, the intervention
has an ICER of <$20,000/HLY gained.
The intervention is also predicted to prevent $40.4 million in medical costs from unwanted pregnancies avoided in
the main scenario (Table 2). In many sensitivity analyses,
the cost avoided from these unwanted pregnancies prevented alone is greater than the intervention cost. The larger the
numbers of no contraception users and less-effective method
users receiving contraceptive services and willing to switch
to more effective methods, the greater the magnitude of cost
savings from unwanted pregnancies avoided (Table 3).
78

Discussion
The results of our modeling analysis suggest that increasing access to effective contraception in the context of the
2016 Zika virus outbreak for women in Puerto Rico who
do not intend to become pregnant could proportionally reduce the number of unintended pregnancies and cases of
ZAM by 25%. The intervention is cost-saving (negative
net cost) when considering the benefits from preventing
ZAM and avoiding Zika virus–exposed pregnancy costs in
the main scenarios and in most of the scenarios we tested.
In scenarios in which the intervention is not cost-saving, it
is still cost-effective relative to accepted cost-effectiveness
thresholds (35). The World Health Organization suggests
that interventions that cost <3 times the gross domestic
product per capita per HLY (equivalent to $150,000 in the
United States and $60,000 in Puerto Rico) are cost-effective and those costing less than gross domestic product per
capita are highly cost-effective (36). When considering additional benefits from preventing unintended pregnancies
not desired at a later time, the intervention is cost-saving in
all scenarios. Previous studies have shown that expanding
access to contraception, especially LARC, is cost-saving
(16,37,38). Likewise, our findings suggest that this intervention could be cost-saving or cost-effective within the
context of a public health emergency response.
Our study has several limitations. First, we project the effects of a hypothetical intervention in place in

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 23, No. 1, January 2017

Contraception during Zika Outbreak, Puerto Rico

Table 3. Sensitivity analyses indicating the number of ZAM cases prevented and Zika virus–associated costs avoided in proposed
intervention to increase access to contraception to women during Zika virus outbreak, Puerto Rico, 2016*
Zika virus–
Total
No. ZAM
Incremental
associated incremental
Cost
Additional cost
cases
intervention cost avoided,
cost,†
per HLY avoided from
Parameter
prevented cost, millions
millions
millions
gained
UP, millions
Main scenario
25
$33.5
$65.2
–$31.7
CS
$40.4
% Women receiving contraceptive services from healthcare provider; main scenario, 50% of no method users, 60% of less-effective
method users, and 100% of moderately effective method users
30% of no method users‡
22
$32.4
$55.8
–$23.5
CS
$34.6
70% of no method users
29
$34.6
$74.5
–$39.9
CS
$46.1
30% of less-effective method users
19
$26.0
$50.0
–$24.0
CS
$31.0
80% of less-effective method users
29
$38.5
$75.2
–$36.8
CS
$46.6
% Women receiving contraceptive services as in
16
$25.2
$40.6
–$15.4
CS
$25.1
NSFG 2011–2013§
% Women willing to change to more effective method;¶ main scenario value: 50%
10%
6
$13.0
$15.8
–$2.8
CS
$9.7
30%
16
$23.2
$40.5
–$17.3
CS
$25.0
80%
39
$48.8
$102.2
–$53.3
CS
$63.3
% Women receiving contraceptive services from
10
$18.2
$25.7
–$7.6
CS
$15.9
healthcare provider as in NSFG 2011–2013 with
30% of them willing to change to a new method
Use of highly effective methods among switchers; main value 50%
67%
27
$38.4
$69.9
–$31.5
CS
$43.3
33%
23
$28.5
$60.4
–$31.8
CS
$37.4
Contraception switching pattern reported in
7
$21.8
$17.0
$4.8
$24,608
$10.5
Colorado Family Planning Initiative#
Dual-method use; 30% of moderately effective method users in main scenario
20% of moderately effective users
24
33.1
61.3
–$28.2
CS
$38.0
50% of moderately effective users
28
34.1
–$38.7
CS
$45.1
72.9
Contraception use distribution at baseline

As reported in 2002 BRFSS survey**
30
33.6
–$44.8
CS
$48.6
78.4
As in Title X clinics in 2014††
14
$30.1
$36.7
–$6.6
CS
$22.7
Rate of ZAM among all live-born infants; main scenario value 58/10,000
32/10,000
14
$33.5
$37.5
–$4.0
CS
$40.4
86/10,000
38
$33.5
$96.3
–$62.8
CS
$40.3
Lifetime costs for microcephaly; main scenario value $3.8 million
$1.9 million
25
$33.5
$33.5
0
CN‡‡
$40.4
$2.2 million
25
$33.5
$39.5
–$6.1
CS
$40.4
$5.5 million
25
$33.5
$93.5
–$60.0
CS
$40.4
Termination of pregnancy with ZAM
20%
25
$33.5
$72.8
–$39.3
CS
$40.4
50%
25
$33.5
$44.1
–$10.6
CS
$40.3
Cost of the program other than providing the contraception at no cost to patients; main scenario value $39/person
$0/person
25
$27.1
$65.2
–$38.0
CS
$40.4
$100/person
25
$43.4
$65.2
–$21.8
CS
$40.4
Annualized LARC device cost
25
$17.5
$65.2
–$47.7
CS
$40.4
Puerto Rico costs§§
25
$30.8
$32.5
–$1.7
CS
$14.4
Discount rate
0%
25
$33.5
$105.4
–$72.0
CS
$40.4
5%
25
$33.5
$52.9
–$19.4
CS
$40.4
*BRFSS, Behavioral Risk Factor Surveillance System; CN, cost-neutral; CS, cost-saving; HLY, healthy life years; LARC, long-acting reversible
contraceptive; NSFG, National Survey of Family Growth; UP, unwanted pregnancy; ZAM, Zika virus–associated microcephaly.
†Total incremental cost is the additional cost of contraception minus Zika virus–associated cost avoided.
‡30% of no contraception users, 60% of less-effective contraceptive method users, 100% of moderately effective contraceptive method users seeking
contraceptive services from healthcare provider during the Zika virus outbreak.
§Based on NSFG 2011–2013, among women of reproductive age who are sexually active, did not intend to become pregnant, and were not using
permanent contraceptive methods, 21% of no contraception users, 33% of less-effective contraceptive method users, 97% of moderately effective
contraceptive method users, and 94% of dual-method users had at least 1 contraceptive service visit in the last 12 months (in total 50%).
¶Based on Title X Family Planning annual report for 2007–2015 in Colorado, 30% of clients who visited Title X clinics switched to a new method.
#Eighteen percentage points of users of moderately effective methods are assumed to switch to highly effective methods, of whom 21% were dualmethod users.
**Contraception distribution in Puerto Rico in 2002 15.9% no method, 41.6% less-effective methods, 40.2% moderately effective methods, and 2.4%
highly effective methods.
††In 2014, in Title X clinics in Puerto Rico, 20% of women at risk for unintended pregnancy used less-effective methods, 77% used moderately effective
methods, and 2% used highly effective methods.
‡‡Intervention cost equals to the medical savings from ZAM cases prevented.
§§Conversion factor of 0.36 applied to pregnancy and ZAM medical costs based on the ratio of per capita medical expenditure in Puerto Rico and in the
United States in 2012 as in Portela et al. 2015 (32); conversion factor of 0.72 applied to costs of supportive care for live-born infants with ZAM, based on
the ratio of annual salary for assistant nurses in Puerto Rico and in the United States (33).

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 23, No. 1, January 2017

RESEARCH

Puerto Rico during the 2016 Zika virus outbreak. However, the qualitative results would apply in future outbreaks. Second, the baseline contraception use distribution is based on a 2002 survey; the current distribution in
Puerto Rico might be different. Third, uncertainty exists
about the effect of the proposed intervention on postintervention contraceptive use distribution; however, the
sensitivity analyses indicate that different distributions of
LARC types among switchers does not have a substantial
influence on the results. Fourth, our study assumes that
women have full access to healthcare providers. In areas
with limited access to providers, the effectiveness of the
intervention might be lower, although Puerto Rico has
a similar ratio of physicians to population as the United
States as a whole (39), and despite a loss of physicians
in recent years, Puerto Rico has a network of providers,
federally qualified health clinics, and Title X providers in
rural and urban areas. Fifth, the distribution of outcomes
of unintended pregnancies in Puerto Rico is uncertain.
We lack data on miscarriage and induced abortion rates in
Puerto Rico and so did not have sufficient data to model
uncertainty in these parameters. The rates of stillbirth and
pregnancy termination among pregnancies with ZAM in
Puerto Rico are also unknown. Our assumed percentage
of live births among pregnancies with recognized ZAM
(65%) compares with a 38% rate reported in French Polynesia during the 2013 Zika virus outbreak (11). Sixth,
pregnancy intentions and use of contraception among
women in Puerto Rico might differ during the Zika virus outbreak compared to preoutbreak periods. Seventh,
our analysis does not consider possibly higher rates of fetal loss and induced abortion among women infected by
Zika virus during early pregnancy or brain abnormalities
or conditions related to Zika virus not involving microcephaly. Eighth, the assumed Zika virus testing costs assume 100% adherence to recommended testing practices;
the actual cost savings taking nonadherence into account
would be lower. Ninth, the cost estimates of ZAM cases in live-born infants do not include costs of managing mental health conditions among parents of affected
infants. Tenth, using private insurance payments might
overstate the healthcare cost of treating ZAM. However,
if the cost of ZAM exceeds $1.9 million, the intervention
is still cost-saving. Finally, if efforts to prevent transmission of Zika virus in Puerto Rico are effective, the rate of
infection in pregnancy and the incidence of ZAM relative
to that projected could be reduced.
Despite its limitations, our study has several strengths.
First, the study is based on the most current available information. Second, the contraception scenarios are based on
real-world programs and have resulted from consultation
with subject matter experts. Third, expenditure data from
a large sample of US residents with commercial health
80

insurance were used to calculate the potential medical cost
of ZAM on the basis of combinations of diagnostic codes
for virus-associated microcephaly, although costs might be
lower for similar children with public insurance. Finally,
sensitivity analyses give consistent results indicating expected net cost savings associated with an intervention that
would increase access to contraception in response to the
Zika virus outbreak in Puerto Rico.
Zika virus can cause devastating birth defects, and infants born with ZAM and their families will require lifelong support. Avoiding unintended pregnancies is a critical intervention to mitigate the effects of ZAM. Efforts to
prevent adverse Zika virus–related pregnancy outcomes in
Puerto Rico are especially important because of limited resources (40). Our analyses suggest that increasing access
to a full range of contraception among women in Puerto
Rico who want to delay or avoid becoming pregnant during a Zika virus outbreak would be a cost-saving strategy
to reduce the effects of ZAM. The magnitude of cost savings is even greater when considering the avoided cost of
unwanted pregnancies prevented.
Acknowledgments
We thank Karen Pazol for providing data for input parameters on
contraception failure rate for dual users and contraception
service use, Martin Meltzer for providing invaluable comments
on revising the manuscript, Hilary Whitham for providing
consultation on TreeAge software and the decision tree structure,
Loretta Gavin for providing data on contraception use in Title X
clinics in Puerto Rico, Margaret (Peggy) Honein for leadership
support, Annelise Arth for consulting on the cost of
microcephaly, Matthew Biggerstaff for reviewing the decision
tree, Alys Adamski for providing the most up-to-date p
ublications related to Zika infection, Christine Olson and Romeo
Galang for providing clinical information on Zika infection
among pregnant women, Carrie Shapiro-Mendoza for reviewing
the first round of analyses and providing feedback, and Howard
Goldberg for consultation on the Puerto Rico contraceptive
use distributions.
Dr. Li is the lead economist in the Division of Reproductive
Health, National Center for Chronic Disease Prevention and
Health Promotion, Centers for Disease Control and Prevention.
Her expertise is health economics and economic evaluation.
References

1. Rasmussen SA, Jamieson DJ, Honein MA, Petersen LR. Zika
virus and birth defects—reviewing the evidence for causality.
N Engl J Med. 2016;374:1981–7. http://dx.doi.org/10.1056/
NEJMsr1604338
2. Tepper NK, Goldberg HI, Bernal MI, Rivera B, Frey MT,
Malave C, et al. Estimating contraceptive needs and increasing access to contraception in response to the Zika virus disease
outbreak—Puerto Rico, 2016. MMWR Morb Mortal Wkly Rep.
2016;65:311–4. http://dx.doi.org/10.15585/mmwr.mm6512e1

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 23, No. 1, January 2017

Contraception during Zika Outbreak, Puerto Rico
3. Oster AM, Russell K, Stryker JE, Friedman A, Kachur RE,
Petersen EE, et al. Update: interim guidance for prevention of
sexual transmission of Zika virus—United States, 2016. MMWR
Morb Mortal Wkly Rep. 2016;65:323–5.
http://dx.doi.org/10.15585/mmwr.mm6512e3
4. Centers for Disease Control and Prevention. Zika virus disease
in the United States, 2015–2016 [cited 2016 May 3].
http://www.cdc.gov/zika/geo/united-states.html
5. Ellington SR, Devine O, Bertolli J, Martinez Quiñones A,
Shapiro-Mendoza CK, Perez-Padilla J, et al. Estimating the number
of pregnant women infected with Zika virus and expected infants
with microcephaly following the Zika outbreak in Puerto Rico,
2016. JAMA Pediatr. 2016 Aug 19 [Epub ahead of print].
6. Finer LB, Zolna MR. Declines in unintended pregnancy in the
United States, 2008–2011. N Engl J Med. 2016;374:843–52.
http://dx.doi.org/10.1056/NEJMsa1506575
7. Trussell J. Overstating the cost savings from contraceptive use.
Eur J Contracept Reprod Health Care. 2008;13:219–21.
http://dx.doi.org/10.1080/13625180802359263
8. Bensyl DM, Iuliano DA, Carter M, Santelli J, Gilbert BC.
Contraceptive use: United States and territories, Behavioral Risk
Factor Surveillance System, 2002. MMWR Surveill Summ.
2005;54:1–72.
9. Secura GM, Allsworth JE, Madden T, Mullersman JL, Peipert JF.
The Contraceptive CHOICE Project: reducing barriers to longacting reversible contraception. Am J Obstet Gynecol.
2010;203:115.e1–7. http://dx.doi.org/10.1016/j.ajog.2010.04.017
10. Trussell J. Contraceptive failure in the United States.
Contraception. 2011;83:397–404. http://dx.doi.org/10.1016/j.
contraception.2011.01.021
11. Sedgh G, Singh S, Henshaw SK, Bankole A. Legal abortion
worldwide in 2008: levels and recent trends. Int Perspect Sex
Reprod Health. 2011;37:84–94. http://dx.doi.org/10.1363/3708411
12. Singh S, Sedgh G, Hussain R. Unintended pregnancy: worldwide
levels, trends, and outcomes. Stud Fam Plann. 2010;41:241–50.
http://dx.doi.org/10.1111/j.1728-4465.2010.00250.x
13. Cragan JD, Gilboa SM. Including prenatal diagnoses in birth
defects monitoring: Experience of the Metropolitan Atlanta
Congenital Defects Program. Birth Defects Res A Clin Mol Teratol.
2009;85:20–9. http://dx.doi.org/10.1002/bdra.20508
14. Centers for Disease Control and Prevention. Outcomes of
pregnancies with laboratory evidence of possible Zika virus
infection in the United States, 2016 [cited 2016 Jul 26].
https://www.cdc.gov/zika/geo/pregnancy-outcomes.html
15. Grosse SD, Ouyang L, Collins JS, Green D, Dean JH,
Stevenson RE. Economic evaluation of a neural tube defect
recurrence-prevention program. Am J Prev Med. 2008;35:572–7.
http://dx.doi.org/10.1016/j.amepre.2008.07.008
16. Trussell J, Hassan F, Lowin J, Law A, Filonenko A. Achieving
cost-neutrality with long-acting reversible contraceptive methods.
Contraception. 2015;91:49–56. http://dx.doi.org/10.1016/
j.contraception.2014.08.011
17. Truven Health Analytics. The cost of having a baby in the United
States [cited 2016 May 16]. http://transform.childbirthconnection.
org/wp-content/uploads/2013/01/Cost-of-Having-a-Baby1.pdf
18. Biggio JR Jr, Morris TC, Owen J, Stringer JS. An outcomes
analysis of five prenatal screening strategies for trisomy 21
in women younger than 35 years. Am J Obstet Gynecol.
2004;190:721–9. http://dx.doi.org/10.1016/j.ajog.2003.09.028
19. Santelli JS, Lindberg LD, Finer LB, Singh S. Explaining
recent declines in adolescent pregnancy in the United States:
the contribution of abstinence and improved contraceptive use.
Am J Public Health. 2007;97:150–6. http://dx.doi.org/10.2105/
AJPH.2006.089169
20. Oduyebo T, Petersen EE, Rasmussen SA, Mead PS,
Meaney-Delman D, Renquist CM, et al. Update: interim guidelines

21.

22.

23.
24.
25.

26.

27.

28.
29.

30.
31.
32.

33.
34.

35.

36.

for health care providers caring for pregnant women and women of
reproductive age with possible Zika virus exposure—United States,
2016. MMWR Morb Mortal Wkly Rep. 2016;65:122–7.
http://dx.doi.org/10.15585/mmwr.mm6505e2
Fleming-Dutra KE, Nelson JM, Fischer M, Staples JE,
Karwowski MP, Mead P, et al. Update: interim guidelines for health
care providers caring for infants and children with possible Zika
virus infection—United States, February 2016. MMWR Morb
Mortal Wkly Rep. 2016;65:182–7. http://dx.doi.org/10.15585/
mmwr.mm6507e1
Martines RB, Bhatnagar J, Keating MK, Silva-Flannery L,
Muehlenbachs A, Gary J, et al. Notes from the field: evidence of
Zika virus infection in brain and placental tissues from two
congenitally infected newborns and two fetal losses—Brazil,
2015. MMWR Morb Mortal Wkly Rep. 2016;65:159–60.
http://dx.doi.org/10.15585/mmwr.mm6506e1
Hamilton BE, Martin JA, Osterman MJ, Curtin SC, Matthews TJ.
Births: Final Data for 2014. Natl Vital Stat Rep. 2015;64:1–64.
Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Menacker F,
Munson ML. Births: final data for 2002. Natl Vital Stat Rep.
2003;52:1–113.
Riehman KS, Sly DF, Soler H, Eberstein IW, Quadagno D,
Harrison DF. Dual-method use among an ethnically diverse
group of women at risk of HIV infection. Fam Plann Perspect.
1998;30:212–7. http://dx.doi.org/10.2307/2991606
Besnard M, Eyrolle-Guignot D, Guillemette-Artur P, Lastère S,
Bost-Bezeaud F, Marcelis L, et al. Congenital cerebral
malformations and dysfunction in fetuses and newborns following
the 2013 to 2014 Zika virus epidemic in French Polynesia. Euro
Surveill. 2016;21:30181. http://dx.doi.org/10.2807/1560-7917.
ES.2016.21.13.30181
Alfaro-Murillo JA, Parpia AS, Fitzpatrick MC, Tamagnan JA,
Medlock J, Ndeffo-Mbah ML, et al. A cost-effectiveness
tool for informing policies on Zika virus control. PLoS Negl
Trop Dis. 2016;10:e0004743. http://dx.doi.org/10.1371/journal.
pntd.0004743
Broyles RS, Tyson JE, Swint JM. Have Medicaid reimbursements
been a credible measure of the cost of pediatric care? Pediatrics.
1997;99:e8. http://dx.doi.org/10.1542/peds.99.3.e8
Trussell J, Lalla AM, Doan QV, Reyes E, Pinto L, Gricar J.
Cost effectiveness of contraceptives in the United States.
Contraception. 2009;79:5–14. http://dx.doi.org/10.1016/
j.contraception.2008.08.003
Arroyo MP. Contraceptive access due to Zika threat. Elnuevodia
[cited 2016 Aug 4]. http://www.elnuevodia.com/english/english/
nota/contraceptiveaccessduetozikathreat-2227239
Fowler C, Gable J, Wang J, Lasater B. Title X family planning
annual report: 2014 national summary. Research Triangle Park
(NC): RTI International; 2015.
Portela M, Sommers BD. On the Outskirts of national health
reform: a comparative assessment of health insurance and
access to care in Puerto Rico and the United States. Milbank Q.
2015;93:584–608. http://dx.doi.org/10.1111/1468-0009.12138
Bureau of Labor Statistics. May 2015 state occupational
employment and wage estimates: Puerto Rico [cited 2016 Jul 20].
http://www.bls.gov/oes/current/oes_pr.htm
Bureau of Economic Analysis. Table 2.5.4. Price indexes for
personal consumption expenditures by function [cited 2016 Aug 5].
http://www.bea.gov/iTable/iTable.cfm?reqid=9&step=3&isuri=1&9
03=69#reqid=9&step=3&isuri=1&903=73
Grosse SD. Assessing cost-effectiveness in healthcare:
history of the $50,000 per QALY threshold. Expert
Rev Pharmacoecon Outcomes Res. 2008;8:165–78.
http://dx.doi.org/10.1586/14737167.8.2.165
McGann PT, Grosse SD, Santos B, de Oliveira V, Bernardino L,
Kassebaum NJ, et al. A cost-effectiveness analysis of a pilot

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 23, No. 1, January 2017

RESEARCH
neonatal screening program for sickle cell anemia in the Republic
of Angola. J Pediatr. 2015;167:1314–9. http://dx.doi.org/10.1016/j.
jpeds.2015.08.068
37. Burlone S, Edelman AB, Caughey AB, Trussell J, Dantas S,
Rodriguez MI. Extending contraceptive coverage under the
Affordable Care Act saves public funds. Contraception. 2013;87:143–
8. http://dx.doi.org/10.1016/j.contraception.2012.06.009
38. Frost JJ, Sonfield A, Zolna MR, Finer LB. Return on investment: a
fuller assessment of the benefits and cost savings of the US publicly funded family planning program. Milbank Q. 2014;92:696–749.
http://dx.doi.org/10.1111/1468-0009.12080

39. Center for Work Force Studies. 2015 state physician workforce
data book. Washington: Association of American Medical
Colleges; 2015.
40. Bishaw A, Fontenot K. Poverty: 2012 and 2013 [cited 2016
May 20]. https://www.census.gov/content/dam/Census/library/
publications/2014/acs/acsbr13-01.pdf
Address for correspondence: Rui Li, Centers for Disease Control and
Prevention, 4770 Buford Hwy NE, Mailstop F-74, Atlanta, GA 30341,
USA; email: [email protected]

August 2014: Vector-borne Diseases
• L eptospirosis-Associated
Hospitalizations, United
States, 1998–2009
• Independent Origin of
Plasmodium falciparum
Antifolate SuperResistance, Uganda,
Tanzania, and Ethiopia
• Global and Local
Persistence of Influenza
A(H5N1) Virus
• Human Exposure to Live
Poultry and Psychological
and Behavioral
Responses to Influenza
A(H7N9), China
• Rapid Whole-Genome
Sequencing for Surveillance
of Salmonella enterica
Serovar Enteritidis

•N
ovel Reassortant
Influenza A(H5N8)
Viruses in Domestic
Ducks, Eastern China
• Antibodies against
MERS Coronavirus in
Dromedary Camels,
Kenya, 1992–2013
•B
orrelia crocidurae
Infection in Acutely
Febrile Patients, Senegal

•S
helter Dogs as Sentinels
for Trypanosoma cruzi
Transmission across
Texas, USA
• Natural Intrauterine
Infection with
Schmallenberg Virus in
Malformed
Newborn Calves

•R
ole of Migratory Birds in
Spreading CrimeanCongo Hemorrhagic
Fever, Turkey
• Isolation of MERS
Coronavirus from
Dromedary Camel,
Qatar, 2014
• New Introductions of
Enterovirus 71
Subgenogroup C4
Strains, France, 2012
• Rapid Detection,
Complete Genome
Sequencing, and
Phylogenetic Analysis of
Porcine Deltacoronavirus

•G
eographic Distribution
of MERS Coronavirus
among Dromedary
Camels, Africa
• Human Infections with
Borrelia miyamotoi,
Japan
• Co-circulation of
Dengue and
Chikungunya Viruses, Al
Hudaydah, Yemen, 2012
• Antibodies against
Severe Fever with
Thrombocytopenia
Syndrome Virus in
Healthy Persons, China,
2013
• Severe Fever with
Thrombocytopenia
Syndrome Virus in Ticks
Collected from Humans,
South Korea, 2013
• Infection with Possible
Precursor of Avian
Influenza A(H7N9) Virus
in a Child, China, 2013

•D
engue Virus
Transmission by Blood
Stem Cell Donor after
Travel to Sri Lanka, 2012
• Severe Murine Typhus
with Pulmonary System
Involvement
• Detection of East/Central/
South African Genotype
of Chikungunya Virus in
Myanmar, 2010
• Pulmonary Infection and
Colonization with
Nontuberculous
Mycobacteria, Taiwan,
2000–2012
• Levofloxacin-Resistant
Haemophilus influenzae,
Taiwan, 2004–2010
• Movement of
Chikungunya Virus into
the Western Hemisphere
• Diagnosis of Bartonella
henselae Prosthetic Valve
Endocarditis in Man,
France

http://wwwnc.cdc.gov/eid/articles/
issue/21/08/table-of-contents
82

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 23, No. 1, January 2017

Research

ajog.org

GYNECOLOGY

Three-year continuation of reversible
contraception
Justin T. Diedrich, MD, MSCI; Qiuhong Zhao, MS; Tessa Madden, MD, MPH;
Gina M. Secura, PhD; Jeffrey F. Peipert, MD, PhD
OBJECTIVE: The objective of this analysis was to estimate the 3-year

RESULTS: Our analytic sample consisted of 4708 CHOICE participants

continuation rates of long-acting reversible contraceptive (LARC)
methods and to compare these rates to non-LARC methods.

who met inclusion criteria. Three-year continuation rates were 69.8%
for users of the levonorgestrel intrauterine device, 69.7% for copper
intrauterine device users, and 56.2% for implant users. At 3 years,
continuation was 67.2% among LARC users and 31.0% among
non-LARC users (P < .001). After adjustment for age, race, education,
socioeconomic status, parity, and history of sexually transmitted
infection, the hazard ratio for risk of discontinuation was 3-fold higher
among non-LARC method users than LARC users (adjusted hazard
ratio, 3.08; 95% confidence interval, 2.80e3.39).

STUDY DESIGN: The Contraceptive CHOICE Project (CHOICE) was a prospective cohort study that followed 9256 participants with telephone surveys at 3 and 6 months, then every 6 months for 2e3 years. We estimated
3-year continuation rates of baseline methods that were chosen at
enrollment. The LARC methods include the 52-mg levonorgestrel intrauterine device; the copper intrauterine device, and the subdermal implant).
These were then compared to rates to non-LARC hormonal methods (depot
medroxyprogesterone acetate, oral contraceptive pills, contraceptive patch,
and vaginal ring). Eligibility criteria for this analysis included participants
who started their baseline chosen method by the 3-month survey. Participants who discontinued their method to attempt conception were
censored. We used a Cox proportional hazard model to adjust for confounding and to estimate the hazard ratio for risk of discontinuation.

CONCLUSION: Three-year continuation of the 2 intrauterine devices

approached 70%. Continuation of LARC methods was significantly
higher than non-LARC methods.
Key words: contraception, continuation, intrauterine device, longacting reversible contraception, subdermal implant

Cite this article as: Diedrich JT, Zhao Q, Madden T, et al. Three-year continuation of reversible contraception. Am J Obstet Gynecol 2015;213:662.e1-8.

ong-acting reversible contraceptive
(LARC) methods are highly effective and have high user satisfaction.1
Their use has increased in the United
States over the past 2 decades; approximately 8.5% of women who use
contraception report current use of a
LARC method.2 In fact, a recent report
demonstrated a nearly 5-fold increase in
LARC methods over the last decade.3
LARC users are likely to be highly satisﬁed with their method at 12 and 24

months.4,5 However, data are lacking
regarding continuation of LARC methods
at 3 years in the United States. Some of the
previous studies that assessed longer-term
continuation randomly assigned women
to a contraceptive method and included
women from many different countries.6,7
The largest study was performed by
Sivin et al. 8 This multinational study
randomly assigned women to the

levonogestrel-20 (the predecessor to the
current levonorgestrel-containing intrauterine device [LNG-IUD]) and the
TCu380Ag (the predecessor to the current
copper-intrauterine device [Cu-IUD]).
Cumulative continuation at 3 years was
49% among LNG-20 users and 59%
among TCu380Ag users. Other prospective studies found 3-year continuation
rates of 67e78% among users of CuIUDs.9,10 Continuation of LNG-IUD has
been reported at 73e80% at 3 years.11,12

From the Divisions of Family Planning and Clinical Research, Department of Obstetrics and Gynecology; Washington University School of Medicine in St.
Louis, St. Louis, MO.
Received April 27, 2015; revised July 20, 2015; accepted Aug. 2, 2015.
The Contraceptive CHOICE Project is funded by an anonymous foundation. This publication also was supported by Washington University Institute of
Clinical and Translational Sciences grant number UL1 TR000448 from the National Center for Advancing Translational Sciences and award number
K23HD070979 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health.
J.F.P. receives research funding/support from Bayer, Teva, and Merck and serves on advisory boards for Teva Pharmaceuticals and MicroCHIPs; T.M.
serves on an advisory board for Bayer Healthcare Pharmaceuticals and a data safety monitoring board for phase 4 safety studies of Bayer contraceptive
products; the remaining authors report no conﬂict of interest.
The content is solely the responsibility of the authors and does not necessarily represent the ofﬁcial views of the National Institutes of Health.
Corresponding author: Jeffrey F. Peipert, MD, PhD. [email protected]
0002-9378/free ª 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ajog.2015.08.001

662.e1 American Journal of Obstetrics & Gynecology NOVEMBER 2015

Gynecology

ajog.org

FIGURE 1

Study inclusion

hypothesized that 36-month continuation rates for the LARC methods would
exceed 60% and that continuation would
be signiﬁcantly higher for LARC
methods than non-LARC methods.

M ATERIALS
Flow chart of participants included in analysis.
Diedrich. Three-year contraceptive continuation. Am J
Obstet Gynecol 2015.

Subdermal implants have international
continuation rates of 30e53%.12-15
This analysis was performed to estimate the rates of 36-month continuation
of the baseline contraceptive method
that was chosen and to compare
continuation rates of LARC and nonLARC methods at enrollment into the
Contraceptive CHOICE Project. In
addition, we explored baseline characteristics that are associated with discontinuation of contraceptive methods. We

AND

M ETHODS

In 2007, the Contraceptive CHOICE
Project (CHOICE) began recruiting
women for a prospective observational
cohort study. The goal of the study was
to reduce the unintended pregnancy rate
in the St. Louis, MO, area by promoting
the most effective methods of contraception and eliminating the cost barrier
to all forms of contraception. The
methods have been reported in detail16
but are described brieﬂy below. The
Human Research Protection Ofﬁce at
Washington University in St. Louis
approved the study protocol before
study recruitment.
Participants were referred to CHOICE
through their health care providers,
posted ﬂyers, and word of mouth.
Recruitment sites included local health

FIGURE 2

Contraceptive continuation

Kaplan-Meier Survival Curve of long-acting reversible contraceptive methods and non-long-acting
reversible contraceptive methods.
LARC, long-acting reversible contraceptive method.
Diedrich. Three-year contraceptive continuation. Am J Obstet Gynecol 2015.

Research

care centers, 2 abortion care providers,
and a university-associated clinical
research center. Inclusion criteria
included women who (1) were 14e45
years of age, (2) desired reversible
contraception and were willing to start a
new method, (3) were sexually active
with a male partner or intended to be
within 6 months, (4) who lived in or
received reproductive care in the St.
Louis area, and (5) were able to consent
in English or Spanish. Women were
excluded if they desired pregnancy in the
next 12 months or were had had hysterectomy or permanent sterilization.
Recruitment of the 9256 participants
began in 2007 and was completed in
2011. All participants provided written
informed
consent
before
study
enrollment.
All potential participants heard a
standardized introduction to LARC
methods; upon enrollment, they
received additional contraceptive counseling.17 LARC methods included the
LNG-IUD, the Cu-IUD, and the 3-year
subdermal implant. The contraceptive
counseling reviewed all reversible methods in order of effectiveness from most
to least effective. After a baseline interview, participants completed screening
for sexually transmitted infections,
received their contraceptive of choice
at no cost, and were followed for 2 or
3 years, depending on the timing of
enrollment. Follow-up telephone interviews were performed at 3 and 6
months then every 6 months thereafter
for the duration of study participation.
At the enrollment visit, each participant
chose her baseline method. When
possible, they would start that method
immediately. In certain cases (such as
when pregnancy could not be ruled out
reasonably), the patient received a
bridge method until they returned for
the initiation of their chosen method.
Bridge methods included depot
medroxyprogesterone acetate (DMPA),
oral contraceptive pills (OCPs), combined contraceptive patch, vaginal ring,
or condoms. Participants were able
to switch methods at any time during the
follow-up period. For the purposes
of this analysis, if a participant switched
her method, we considered this a

NOVEMBER 2015 American Journal of Obstetrics & Gynecology

662.e2

Research

Gynecology

ajog.org

TABLE 1

Baseline characteristics of analytic sample, stratified by contraceptive method and age
Variable
a

Age, y

Overall
(n [ 4708)

Non-long-acting reversible
contraceptive method
(n [ 1505)

Long-acting reversible
contraceptive method
(n [ 3203)

P value

25.2 5.7

24.0 5.0

25.7 5.9

< .001

Race, n (%)

.668

Black

2243 (47.7)

723 (48.1)

1520 (47.5)

White

2100 (44.6)

659 (43.8)

1441 (45.0)

Other

364 (7.7)

122 (8.1)

242 (7.6)
< .001

Education, n (%)
High school

1661 (35.3)

472 (31.4)

1189 (37.1)

Some college

1987 (42.2)

666 (44.3)

1321 (41.3)

College graduate

1058 (22.5)

366 (24.3)

692 (21.6)
< .001

Body mass index, n (%)
Underweight

144 (3.1)

72 (4.9)

72 (2.3)

Normal

1895 (41.2)

718 (49.2)

1177 (37.4)

Overweight

1206 (26.2)

333 (22.8)

873 (27.8)

1357 (29.5)

336 (23.0)

1021 (32.5)

Obese
Low socioeconomic status, n (%)

< .001

2088 (44.4)

776 (51.6)

1312 (41.0)

Yes

2618 (55.6)

728 (48.4)

1890 (59.0)
< .001

Insurance, n (%)
None

2031 (43.5)

683 (46.0)

1348 (42.4)

Private

2081 (44.6)

711 (47.9)

1370 (43.0)

Public

556 (11.9)

91 (6.1)

465 (14.6)

2225 (47.3)

988 (65.6)

1237 (38.7)

1150 (24.4)

290 (19.3)

860 (26.8)

819 (17.4)

149 (9.9)

670 (20.9)

3þ

514 (10.9)

78 (5.2)

436 (13.6)

< .001

Parity, n (%)

< .001

Unintended pregnancies, n (%)
0

1599 (34.0)

704 (46.9)

895 (28.0)

1292 (27.5)

423 (28.2)

869 (27.2)

780 (16.6)

188 (12.5)

592 (18.5)

1027 (21.9)

187 (12.5)

840 (26.3)

3þ

Diedrich. Three-year contraceptive continuation. Am J Obstet Gynecol 2015.

discontinuation of the baseline method.
Women who received the implant
were told that it was approved for up
to 3 years of use. If a participant had
the device removed and reinserted
within the same month, it was not

considered a discontinuation. Follow-up
interviews focused on method use,
complaints, complications, side-effects,
method troubleshooting, reasons for
method discontinuation, and pregnancies. CHOICE participants have

662.e3 American Journal of Obstetrics & Gynecology NOVEMBER 2015

(continued)

unrestricted access to device removal,
even after CHOICE ended.
This analysis included women who
chose a LARC or a non-LARC method
(DMPA, OCPs, contraceptive patch, or
vaginal ring), started using their method

Gynecology

ajog.org

Research

TABLE 1

Baseline characteristics of analytic sample, stratified by contraceptive method and age (continued)
Overall
(n [ 4708)

Variable

Non-long-acting reversible
contraceptive method
(n [ 1505)

Long-acting reversible
contraceptive method
(n [ 3203)

P value
< .001

History of abortion at baseline, n (%)
No

2870 (61.0)

969 (64.4)

1901 (59.4)

Yes

1838 (39.0)

536 (35.6)

1302 (40.6)

History of sexually transmitted
infection at baseline

.011

No
Yes
a

2869 (61.0)

957 (63.6)

1912 (59.7)

1836 (39.0)

547 (36.4)

1289 (40.3)

Data are given as mean SD; Includes trouble paying for basic necessities or receiving government subsidies in the form of food stamps or welfare.
b

Diedrich. Three-year contraceptive continuation. Am J Obstet Gynecol 2015.

TABLE 2

Kaplan-Meier estimates of 1-, 2-, and 3-year continuation of baseline
method chosen
Continuation, % (95% confidence interval)
Variable

1 Year

2 Year

3 Year

Overall

76.7 (75.4e77.9)

64.2 (62.6e65.5)

56.2 (54.5e57.5)

Levonorgestrel

87.3 (85.8e88.6)

76.7 (74.8e78.5)

69.8 (67.6e71.8)

Copper

84.3 (80.7e87.3)

76.2 (72.1e79.9)

69.7 (65.1e73.7)

Implant

81.7 (78.3e84.7)

68.7 (64.7e72.3)

56.2 (51.8e60.3)

Depot medroxyprogesterone
acetate

57.1 (51.6e62.3)

39.3 (33.8e44.7)

33.2 (26.9e37.7)

Oral contraceptive pill

60.6 (56.3e64.6)

42.2 (37.9e46.5)

31.5 (27.3e35.8)

Ring

54.3 (49.7e58.6)

37.5 (33.1e41.9)

30.0 (25.8e34.4)

Patch

48.2 (38.3e57.4)

35.0 (25.7e44.5)

28.4 (19.5e37.9)

Long-acting reversible
contraceptive

85.8 (84.5e87.0)

75.2 (73.6e76.7)

67.2 (65.4e68.9)

Non-long-acting reversible
contraceptive

55.8 (54.2e59.4)

39.5 (36.9e42.1)

31.0 (28.5e33.5)

Long-acting reversible
contraceptive

82.1 (78.0e85.6)

68.0 (63.0e72.5)

52.6 (47.2e57.7)

Non-long-acting
reversible contraceptive

48.5 (42.1e54.6)

34.5 (28.5e40.6)

23.1 (17.6e29.0)

Long-acting reversible
contraceptive

86.3 (85.0e87.6)

76.2 (74.5e77.8)

69.2 (67.4e71.0)

Non-long-acting
reversible contraceptive

58.6 (55.7e61.3)

40.5 (37.7e43.4)

32.6 (29.8e35.4)

Intrauterine device

Adolescents, 14e19 y

Adults, 20e45 y

Diedrich. Three-year contraceptive continuation. Am J Obstet Gynecol 2015.

by their 3-month survey, and completed
their 36-month follow-up survey or had
another data source that veriﬁed
continuation or discontinuation at 3
years. Continuation rates at 3 years were
estimated for each method. LARC
methods were compared with nonLARC methods and were stratiﬁed by
age (14e19 and 20e45 years old).
Descriptive analyses were performed to
describe demographic characteristics of
participants with the use of chi-square
test or t-test, where appropriate.
Normality was assessed for continuous
variables. The time-to-event for this
survival analysis was calculated from
method initiation to the time point when
the participant discontinued her contraceptive method. If she was lost to
follow up, she was censored at her last
time of contact with CHOICE. Participants were censored if they discontinued
a contraceptive method to attempt
pregnancy. Kaplan-Meier survival functions were used to estimate continuation
rates among different methods. We used
Cox proportional hazard models to estimate hazard ratios for risk of contraceptive method discontinuation for
characteristics that were associated with
discontinuation. We deﬁned confounders
as variables that changed the estimate of
hazard ratio for a contraceptive method
by 10% when they were included in the
model. Confounding variables and signiﬁcant factors from univariable analysis
or variables that were set a priori were

NOVEMBER 2015 American Journal of Obstetrics & Gynecology

662.e4

Research

Gynecology

included in the ﬁnal multivariable model
to evaluate their effect size. The alpha
level was set at .05. Stata software
(version 11; StataCorp, College Station,
TX) was used for all analyses.

R ESULTS
Of 9256 CHOICE participants, the ﬁrst
5090 were observed for 3 years. In this
cohort, 382 women were excluded
because they did not start their chosen
baseline method by the 3-month survey.
There were a total of 4708 participants
(92%) who were observed for 3 years and
were included in this analysis. A ﬂow
diagram of included participants is
shown in Figure 1. There were 185
women (4%) who were censored
because of discontinuation for desire to
conceive or pregnancy.
Demographic and reproductive characteristics are shown in Table 1. Mean
age of participants in this analysis was 25
years; 48% were black; 35% had a high
school education or less; 34% received
public assistance; 44% had no health
insurance; and 12% reported public insurance. Overall, 47% were nulliparous,
and 66% reported at least 1 unintended
pregnancy at baseline. There were 662
adolescents in our cohort: 405 used
LARC methods and 257 used non-LARC
methods. In our stratiﬁed analysis by
LARC vs non-LARC methods, we noted
that LARC users were older, had higher
parity, were more likely to have public
insurance, and were more likely to have a
history of an unintended pregnancy.
Continuation at 1, 2, and 3 years for
each contraceptive method is listed in
Table 2. We stratiﬁed continuation by
LARC and non-LARC methods. At 3
years, continuation was 67.2% among
LARC users and 31.0% among nonLARC users (P < .001) (Figure 2). The
highest continuation was among IUD
users, with 69.8% continuation among
LNG-IUD users and 69.7% among CuIUD users. Non-LARC methods had
lower rates of continuation that range
from 28e33% at 3 years. Among adolescents 14e19 years old, 3-year continuation was lower for all methods
compared with women 20e45 years old
and was lowest among non-LARC
methods (52.6% for adolescents who

ajog.org

TABLE 3

Univariable analysis of risk factors for discontinuation of baseline
contraceptive method at 3 years
Univariable model
Variable

Hazard ratio

95% Confidence interval

Contraceptive method
Oral contraceptive pill

Reference

Intrauterine device
Levonorgestrel

0.30

0.27e0.35

Copper

0.31

0.26e0.38

Implant

0.48

0.40e0.56

Depot medroxyprogesterone acetate

1.01

0.85e1.20

Patch

1.21

0.94e1.56

Ring

1.12

0.96e1.30

1.55

1.38e1.74

Age, y
14e19
20þ

Reference

Race
Black

1.20

1.09e1.32

1.30

1.10e1.54

White

Reference

Other
Education
High school

Reference

Some college

0.93

0.84e1.03

College graduate

0.82

0.73e0.93

1.18

0.92e1.51

0.91

0.82e1.02

0.78

0.70e0.87

Body mass index
Underweight
Normal

Reference

Overweight
Obese
Low socioeconomic status

No
Yes

Reference
0.95

0.87e1.04

1.07

0.97e1.18

Insurance
None
Commercial
Public

Reference
1.04

0.90e1.20

1.38

1.27e1.51

Parity
0
1þ
Diedrich. Three-year contraceptive continuation. Am J Obstet Gynecol 2015.

662.e5 American Journal of Obstetrics & Gynecology NOVEMBER 2015

Reference
(continued)

Gynecology

ajog.org

TABLE 3

Univariable analysis of risk factors for discontinuation of baseline
contraceptive method at 3 years (continued)
Univariable model
Variable

Hazard ratio

95% Confidence interval

0.79

0.72e0.87

Previous unintended pregnancies
0

Reference

1þ
History of sexually transmitted infection
No

Reference

Yes
a

1.13

1.03e1.24

Defined as trouble paying for basic needs (food, housing, medical care, transportation) or receiving government aid (food
stamps, welfare).

Diedrich. Three-year contraceptive continuation. Am J Obstet Gynecol 2015.

TABLE 4

Multivariable analysis of risk factors for discontinuation of baseline
contraceptive method at 3 years
Multivariable model
Variable

Hazard ratio

95% Confidence interval

Contraceptive method
Oral contraceptive pills

Reference

Intrauterine device
Levonorgestrel

0.31

0.27e0.36

Copper

0.33

0.27e0.40

Implant

0.44

0.37e0.52

Depot medroxyprogesterone acetate

0.93

0.78e1.11

Patch

1.17

0. 91e1.52

Ring

1.16

1.00e1.35

Contraceptive duration
Long-acting reversible contraceptive
Non-long-acting reversible contraceptive

Reference
3.08

2.80e3.39

1.33

1.16e1.53

Age, y
14e19
20þ

Reference

Race
Black

1.12

White
Other

1.01e1.25
Reference

1.26

1.07e1.50

Education
High school

used LARC and 23.1% for non-LARC
methods). By 3 years, 54.6% of adolescents continued the LNG-IUD; 49.5%
continued the Cu-IUD, and 50.8%
continued the subdermal implant.
Univariable analysis of risk factors for
discontinuation is shown in Table 3; the
multivariable model is shown in Table 4.
After adjustment for age, race, education,
low socioeconomic status, parity, and history of sexually transmitted infection, the
hazard ratio for discontinuation was >3
times higher among non-LARC method
users (adjusted hazard ratio, 3.08; 95%
conﬁdence interval, 2.80e3.39) than
LARC users. Participants 14-19 years old at
baseline were more likely to discontinue at
3 years compared with women 20 years
old (adjusted hazard ratio, 1.33; 95%
conﬁdence
interval,
1.16e1.53).
Compared to those with a high school
education or less, college graduates reported a lower risk of discontinuation
(adjusted hazard ratio, 0.85; 95% conﬁdence interval, 0.74e0.98).
Participants discontinued their baseline methods for a variety of reasons
(Table 5). Of LNG-IUD users who discontinued this method, approximately
19% did so because of bleeding changes;
25% reported “I did not like how it made
me feel.” The most common reasons for
Cu-IUD users to stop their method was
bleeding changes (35%) and cramping
(17%). Forty-ﬁve percent of implant
discontinuers
reported
bleeding
changes; 28% reported that they did not
like how they felt. Among non-LARC
methods, 33% of DMPA users who
stopped this method reported general
side-effects as the most common reason
for discontinuation. Forty-two percent
of OCP users who discontinued reported
logistical reasons, such as the pill being
hard to remember to take or hard to get.
Of patch discontinuers, 41% reported
side-effects. Twenty-seven percent of
women who discontinued the ring reported side-effects; 24% reported logistical issues.

Reference

C OMMENT
Continuation rates for LARC methods at
1, 2, and 3 years are signiﬁcantly higher
than non-LARC methods. After adjustment for confounding variables, the

Some college

0.93

0.84e1.04

College/graduate

0.85

0.74e0.98

Diedrich. Three-year contraceptive continuation. Am J Obstet Gynecol 2015.

Research

(continued)

NOVEMBER 2015 American Journal of Obstetrics & Gynecology

662.e6

Research

Gynecology

ajog.org

TABLE 4

Multivariable analysis of risk factors for discontinuation of baseline
contraceptive method at 3 years (continued)
Multivariable model
Variable
Low socioeconomic status

Hazard ratio

95% Confidence interval

1.05

0.95e1.17

1.10

0.98e1.23

Reference

Yes
Parity
0
1þ

Reference

History of sexually transmitted disease
No

Reference

Yes
a

1.21

1.10e1.34

Low socioeconomic status is defined as trouble paying for basic needs (food, housing, medical care, transportation) or
receiving government aid (food stamps, welfare)

Diedrich. Three-year contraceptive continuation. Am J Obstet Gynecol 2015.

choice of a shorter-acting method and
younger age were associated with
increased discontinuation.1,4,18,19 Teens
14e19 years of age were more likely to
discontinue than women of 20 years
old. However, of those adolescents who
chose LARC methods, more than onehalf were still using their method at 3

years compared with one-ﬁfth of adolescents who were using their non-LARC
methods at 3 years. Even among women
who were using short-acting methods, 3year continuation was relatively high.
Those who used OCPs, DMPA, patch,
and ring had continuation rates of
28e33%.

Relatively few previous studies have
estimated continuation beyond 1 year
among women in the United States who
chose their contraceptive method rather
than being randomly assigned to a
method. Furthermore, this analysis was
conducted with the data from the
CHOICE cohort with a high follow-up
rate in which only 19% of participants
were lost to follow up at 3 years. Our
cohort had continuation rates that were
consistent with other prospective trials.
The 70% continuation of both LNG- and
Cu-IUDs was consistent with the previously reported 67e80%.9-12 The 56%
continuation of implants was higher than
the previously reported 30e53%.12-15
It is very plausible that women who are
interested in long-term (>2 years) protection from pregnancy are more likely to
select IUDs and the implant; women who
are less certain about their need or desire
for long-term contraception would select
non-LARC methods. However, we still
believe our estimates of 3-year continuation are important for contraceptive
counseling. High continuation rates reﬂect
high satisfaction with LARC methods.5
The major limitation of CHOICE is
that it is a convenience sample within 1
geographic region. Participants were
required to start or switch to a different

TABLE 5

Reasons for discontinuation of baseline contraceptive method in Contraceptive CHOICE Project
Long-acting reversible contraception, n (%)

Non-long-acting reversible contraception, n (%)

Variable

Levonorgestrelintrauterine
Copper-intrauterine
device
device
Implant

Oral
Depo
medroxyprogesterone contraceptive
pill
Patch
acetate

Bleeding changes

136 (19.1)

75 (35.2)

Pain
Did not like “side-effects”
Desired pregnancy

58 (25.6)

49 (12.3)

4 (4.8)

39 (10.4)

2 (2.4)

9 (2.4)

82 (11.5)

37 (17.4)

8 (2.7)

4 (1.8)

3 (0.8)

181 (25.4)

20 (9.4)

81 (27.7)

76 (33.5)

62 (15.6)

67 (9.4)

22 (10.3)

17 (5.8)

4 (1.8)

8 (2.0)

3 (3.6)

15 (4.0)

9 (1.3)

3 (1.4)

2 (0.9)

25 (6.3)

4 (4.8)

11 (2.9)

96 (13.5)

26 (12.2)

—

5 (6.0)

14 (3.7)

Method failed
Expulsion/came off /fell out

133 (45.5)

Ring

—

Difficult to use

Logisticsa

22 (9.7)

169 (42.5)

14 (16.9)

89 (23.7)

61 (26.7)

82 (20.6)

17 (20.5)

78 (20.8)

All others

142 (19.8)

TOTALS

713

30 (14.1)
213

Time, hard to get, remember.

Diedrich. Three-year contraceptive continuation. Am J Obstet Gynecol 2015.

662.e7 American Journal of Obstetrics & Gynecology NOVEMBER 2015

53 (18.2)
292

227

34 (41.0) 100 (26.7)

398

20 (5.3)

375

Gynecology

ajog.org
contraceptive method at enrollment,
which may also limit generalizability. An
important consideration in this analysis
is our assessment of 3-year continuation
of the subdermal implant. This may be
an inappropriate cut-off to measure
continuation and satisfaction if women
undergo implant removal because of the
approaching “expiration” of the device
at 3 years (the Food and Drug Administration-approved duration of use).20
Another limitation is that data on sideeffects, continuation, and expulsion
were self-reported through phone follow
up instead of at clinic visits.
Regardless of age, sociodemographic
markers, and education, women who use
LARC methods report high continuation
rates at 3 years. OCPs and condoms
continue to be the reversible contraceptive methods most commonly used by
women in the United States.3 It is time
for a paradigm shift: LARC methods
should be considered ﬁrst-line contraceptives for women of all ages, given that
satisfaction, continuation, and effectiveness have been shown to be superior
to non-LARC methods.1,4,5,21
-

REFERENCES
1. Winner B, Peipert JF, Zhao Q, et al. Effectiveness of long-acting reversible contraception.
N Engl J Med 2012;366:1998-2007.
2. Daniels K, Daugherty J, Jones J. Current
contraceptive status among women aged 15
-44: United States, 2011 -2013. NCHS data
brief, no 173. Hyattsville, MD: National Center for
Health Statistics. 2014.

3. Branum A, Jones J. Trends in long-acting
reversible contracepiton use among U.S.
women aged 15-44. NCHS Data Brief
2015;188:1-8.
4. O’Neil-Callahan M, Peipert JF, Zhao Q,
Madden T, Secura G. Twenty-four-month
continuation of reversible contraception. Obstet
Gynecol 2013;122:1083-91.
5. Peipert JF, Zhao Q, Allsworth JE, et al.
Continuation and satisfaction of reversible
contraception. Obstet Gynecol 2011;117:
1105-13.
6. Sivin I, Alvarez F, Diaz J, et al. Intrauterine
contraception with copper and with levonorgestrel: a randomized study of the TCu 380Ag
and levonorgestrel 20 mcg/day devices.
Contraception 1984;30:443-56.
7. Andersson K, Odlind V, Rybo G. Levonorgestrel-releasing and copper-releasing (Nova T)
IUDs during ﬁve years of use: a randomized
comparative trial. Contraception 1994;49:
56-72.
8. Sivin I, el Mahgoub S, McCarthy T, et al.
Long-term contraception with the levonorgestrel 20 mcg/day (LNg 20) and the copper T
380Ag intrauterine devices: a ﬁve-year randomized study. Contraception 1990;42:
361-78.
9. UNDP/UNFPA/WHO/World
Bank
IRG.
A randomized multicentre trial of the Multiload
375 and TCu380A IUDs in parous women:
three-year results. UNDP/UNFPA/WHO/World
Bank, Special Programme of Research, Development and Research Training in Human
Reproduction: IUD Research Group. Contraception 1994;49:543-9.
10. Champion CB, Behlilovic B, Arosemena JM,
Randic L, Cole LP, Wilkens LR. A three-year
evaluation of TCu 380 Ag and multiload Cu 375
intrauterine devices. Contraception 1988;38:
631-9.
11. Baldaszti E, Wimmer-Puchinger B,
Löschke K. Acceptability of the long-term contraceptive levonorgestrel-releasing intrauterine
system (Mirena): a 3-year follow-up study.
Contraception 2003;67:87-91.

Research

12. Weisberg E, Bateson D, McGeechan K,
Mohapatra L. A three-year comparative study of
continuation rates, bleeding patterns and satisfaction in Australian women using a subdermal
contraceptive implant or progestogen releasingintrauterine system. Eur J Contracept Reprod
Health Care 2014;19:5-14.
13. Lakha F, Glasier AF. Continuation rates of
Implanon in the UK: data from an observational
study in a clinical setting. Contraception
2006;74:287-9.
14. Rai K, Gupta S, Cotter S. Experience with
Implanon in a north-east London family planning
clinic. Eur J Contracept Reprod Heal Care
2004;9:39-46.
15. Agrawal A, Robinson C. An assessment of
the ﬁrst 3 years’ use of Implanon in Luton. J Fam
Plan Reprod Health Care 2005;31:310-2.
16. Secura GM, Allsworth JE, Madden T,
Mullersman JL, Peipert JF. The Contraceptive
CHOICE Project: reducing barriers to longacting reversible contraception. Am J Obstet
Gynecol 2010;203:115.e1-7.
17. Madden T, Mullersman JL, Omvig KJ,
Secura GM, Peipert JF. Structured contraceptive
counseling provided by the Contraceptive
CHOICE Project. Contraception 2013;88:243-9.
18. Rosenstock JR, Peipert JF, Madden T,
Zhao Q, Secura GM. Continuation of reversible
contraception in teenagers and young women.
Obstet Gynecol 2012;120:1298-305.
19. Secura GM, McNicholas C. Long-acting
reversible contraceptive use among teens prevents unintended pregnancy: a look at the evidence. Expert Rev Obstet Gynecol 2013;8:
297-9.
20. McNicholas C, Maddipati R, Zhao Q,
Swor E, Peipert JF. Use of the etonogestrel
implant and levonorgestrel intrauterine device
beyond the U.S. Food and Drug Administrationapproved duration. Obstet Gynecol 2015;125:
599-604.
21. Secura GM, Madden T, McNicholas C, et al.
Provision of no-cost, long-acting contraception
and teenage pregnancy. N Engl J Med
2014;371:1316-23.

NOVEMBER 2015 American Journal of Obstetrics & Gynecology

662.e8

NIH Public Access
Author Manuscript
J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

NIH-PA Author Manuscript

Published in final edited form as:
J Adolesc Health. 2013 January ; 52(1): 77–82. doi:10.1016/j.jadohealth.2012.08.001.

With Pills, Patches, Rings, and Shots: Who Still Uses Condoms?
A Longitudinal Cohort Study
Rachel L. Goldstein, M.D.a, Ushma D. Upadhyay, Ph.D., M.P.H.b, and Tina R. Raine, M.D.,
M.P.H.c,*
aDepartment of Pediatrics, Stanford University School of Medicine, Palo Alto, California
bAdvancing

New Standards in Reproductive Health and the Bixby Center for Global Reproductive
Health, University of California, San Francisco
cWomen’s

Health Research Institute, at Kaiser Permanente Northern California (Oakland),
formerly with the Bixby Center for Global Reproductive Health, Department of Obstetrics,
Gynecology and Reproductive Sciences, University of California, San Francisco

NIH-PA Author Manuscript

Abstract
Purpose—To describe women’s condom use patterns over time and assess predictors of dual
method use 12 months after initiating hormonal contraceptives.
Methods—We conducted a prospective cohort study among women aged 15–24 years initiating
oral contraceptive pills, patch, ring, or depot medroxyprogesterone and attending public family
planning clinics. Participants completed questionnaires at baseline and 3, 6, and 12 months after
enrollment. We used multivariable logistic regression to assess baseline factors associated with
dual method use at 12 months among 1,194 women who were sexually active in the past 30 days.

NIH-PA Author Manuscript

Results—At baseline, 36% were condom users, and only 5% were dual method users. After
initiation of a hormonal method, condom use decreased to 27% and remained relatively
unchanged thereafter. Dual method use increased to a peak of 20% at 3 months but decreased over
time. Women who were condom users at baseline had nearly twice the odds of being a dual
method user at 12 months compared with nonusers (adjusted odds ratio [AOR] = 2.01, 95% CI:
1.28–3.14). Women who believed their main partner thought condoms were “very important,”
regardless of perceived sexually transmitted infection risk or participant’s own views of condoms,
had higher odds of dual method use (AOR = 2.89, 95% CI: 1.47–5.71).
Conclusions—These results highlight a potential missed opportunity for family planning
providers. Providers focus on helping women initiate hormonal methods, however, they may
improve outcomes by giving greater attention to method continuation and contingency planning in
the event of method discontinuation and to the role of the partner in family planning.
Keywords
Dual method; Contraception; Adolescent; Young adult; Condoms; Sexually transmitted diseases;
Depot medroxyprogesterone acetate; Oral contraceptives; Contraceptive patch; Vaginal ring
Between 1996 and 2006, pregnancy rates among teenaged women (aged 15–19 years) in the
United States decreased by nearly 33%, yet as of 2006 (the most recent year for which data

© 2013 Society for Adolescent Health and Medicine. All rights reserved.
*
Address correspondence to: Tina R. Raine, M.D., M.P.H., Women’s Health Research Institute, Division of Research, Kaiser
Permanente Northern California, 2101 Webster St., 20th Floor, Oakland, CA 94612. [email protected] (T.R. Raine).

Goldstein et al.

Page 2

NIH-PA Author Manuscript

are available) an estimated 82% of these were unintended [1–3]. Sexually experienced teens
and young adults have unintended pregnancy rates more than twice the national figure (69
per 1,000) for sexually active women of childbearing age, with the highest rates (162 per
1,000) among 18–19 year olds [4]. In addition to disproportionately high rates of unintended
pregnancy, women aged 15–24 years also experience high rates of sexually transmitted
infections (STIs). Although comprising only 25% of the sexually active population, teens
and young adults are responsible for more than half of gonorrhea infections and nearly 75%
of chlamydia infections [5]. Unintended pregnancy and STIs remain high, despite
widespread use of contraceptives. Between 2006 and 2010, more than 86% of never-married
female teens and 93% of never-married male teens had used a contraceptive method at last
sex. Of this, condom use accounted for 75% and 52% by men and women, respectively [1].

NIH-PA Author Manuscript

Dual method use, defined as the use of a contraceptive method plus condoms, has been
promoted as an effective way to mitigate the burden of both unintended pregnancy and STIs
in teens and young adults. Although the prevalence of dual method use among teens has
been found to be as high as 20%, when young adults are included, dual method use is as low
as 8.3% [1,6]. Our understanding of factors associated with increased dual method use is
very limited. Much of the prior research on dual method use has been cross-sectional, which
is inadequate to assess the temporal relationship between factors that might contribute to
dual method use [7–10]. In addition, few studies have been designed to analyze continued
condom use at the time that women initiate hormonal contraception. Finally, many prior
studies have been hampered by methodological flaws, including small sample sizes [7,9–
12], differing lengths of follow-up for women using different methods [12], inconsistent
definitions of dual method use including ineffective methods such as abstinence or
withdrawal [7], and inconsistent definitions of discontinuation (discontinuation of condoms
or the hormonal method), limiting the inferences to be drawn and generalizability of this
prior research [13].
We examined condom use patterns over time in a large cohort of high-risk young women
initiating hormonal contraception, including relatively newer contraceptive methods not
previously studied: the transdermal patch and the vaginal ring. Additionally, we sought to
identify predictors of dual method use over a 1-year period to inform strategies for
increasing dual method use.

Materials and Methods
Subjects

NIH-PA Author Manuscript

Data for this study were collected as part of a larger study on factors associated with method
discontinuation and pregnancy among adolescents and young women initiating hormonal
contraception; detailed description of the study methods are described elsewhere [14]. In the
original cohort, women initiating hormonal contraceptives, either the pill, patch, ring, or
depot medroxyprogesterone, were recruited from four Planned Parenthood clinics in
Northern California (Vallejo, Richmond, East Oakland, and Hayward) between September
2005 and July 2008. The study was designed specifically to examine newer short acting
hormonal method use and therefore women using other effective methods including long
acting reversible contraceptive (LARC) methods at baseline were not enrolled. Women,
however, could switch to any method including LARCs over the 1-year follow-up period,
although few did this. Women who presented for reproductive health care were screened
consecutively. Eligibility criteria included being between 15 and 24 years old, not married,
able to read English or Spanish, not pregnant (self-report) or desiring pregnancy within the
next year, and able to provide written informed consent and comply with study procedures.
Women could not have previously used the method they were initiating at the visit. Research
staff collected data from enrolled participants via self-administered electronic questionnaires
J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

Goldstein et al.

Page 3

NIH-PA Author Manuscript

at baseline and 3, 6, and 12 months. All participants provided written informed consent.
Given that minors can consent to contraceptive services in California without parental
consent and that attempting to obtain parental consent could have compromised the
adolescents’ guarantee of confidential services, parental consent was not required. The study
was approved by the Committee on Human Research at the University of California, San
Francisco.
Measures
The primary outcome measure was dual method use at 1 year. Dual method use was defined
as condom use plus an effective contraceptive method. Effective contraceptive methods
included the pill, patch, ring, implant, or IUD. Effective contraceptive method use was
determined from questions about method used at last sex and continued use of the hormonal
method initiated at baseline. Women were considered condom users if the percent of time
they reported using a condom divided by the number of times they reported having sex in
the past 30 days was equal to or greater than 80%. We based our definition of a condom user
on evidence from a recent cohort study that demonstrated that using a proportion of
protected acts (number of times a condom was used divided by the number of vaginal sex
acts during a typical month in the past 3 months) was more predictive of pregnancy
incidence than other measures (since last visit, at last sex, or frequency measure), although
no one method was most predictive of STI/HIV incidence [15].

NIH-PA Author Manuscript

Independent variables considered for the analysis included those found to be associated with
dual method use in previous studies as well as variables informed by the Health Belief
Model, which states that individuals weigh the costs and benefits of a health-related
behavior before attempting behavior change, and the Theory of Planned Behavior, which
takes into account subjective norms around the behavior based on attitudes of individuals
close to the person [16,17]. We grouped variables into the following categories:
sociodemographic characteristics, reproductive history, and attitudes toward condom use.

NIH-PA Author Manuscript

Sociodemographic characteristics collected at baseline included age, neighborhood income,
race and ethnicity, education, and employment status. Sexual and reproductive history
measures included prior pregnancies and STIs, partner concurrency (having sex with a man
other than main partner), and length of time they had sex with their main partner (0–3
months, 4–6 months, 7–12 months, >12 months, and no sex yet). Perceived STI risk in the
next 3 months was measured on a Likert scale (not at all likely, a little likely, somewhat
likely, very likely, don’t know). Participant’s beliefs toward condom use were derived from
responses to a series of questions with answers on a Likert scale. The items were “Condoms
should always be used, even if the girl uses birth control like the pill, patch, ring, or the
shot” and “A girl does not need to use condoms if she gets checked at the clinic often”
(responses for both questions included: strongly agree, agree, neither, disagree, and strongly
disagree). Assessment of the attitudes of the woman and her partner toward condoms was
obtained from the following questions: “How important do you think it is for your main
partner to use condoms when he has sex with you? FOR HIM is it…” and “How important
is it for YOU to use condoms when you have sex with your main partner?” (responses
included: not at all important, somewhat important, very important, and don’t know). For
women who reported condom use at last sex, reason for condom use was also asked
(responses included: STI prevention, pregnancy prevention, both, or don’t know).
Data analysis
Analysis was limited to the subset of women from the original study cohort who reported
having sex in the past 30 days at baseline. Women were divided into those who were
condom users at baseline and those that were not. Comparisons between condom users and

J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

Goldstein et al.

Page 4

NIH-PA Author Manuscript

non-condom users at baseline were made using chi-square analyses. Bivariate analyses (chisquare) were conducted using sociodemographic, reproductive history, and attitude variables
at baseline to model dual method use at 12 months. Multivariable logistic regression was
used to examine factors associated with dual method use at 12 months. Variables chosen for
the multivariable model were based on results from bivariate analyses (p < .05), potential
confounders, prior research, and Health Belief Model/Theory of Planned Behavior. Attrition
analyses were conducted comparing baseline characteristics including sociodemographics
and reproductive history between those lost to follow-up and those who remained in the
study. Two separate sensitivity analyses were done for the multivariable model, the first
assuming those lost to follow-up were dual method users and the second assuming they were
not. Statistical significance level was set to p < .05. All analyses were conducted using
STATA 11 (Stata Corporation, College Station, TX).

Results

NIH-PA Author Manuscript

Of the 1,387 women enrolled at baseline in the cohort, we excluded 193 subjects for the
following reasons: 134 women had not had sex in the past 30 days and 59 women were
missing data on predictor variables. This resulted in 1,194 women who were eligible for
analysis. The cohort was racially/ethnically diverse, with 61% describing themselves as
either Latina or African-American. Nearly two thirds of women were ages 15 to 19 and
more than half lived in a low-income neighborhood. At baseline, 36% of women were
condom users and 5% were dual method users. Condom users at baseline were more likely
to have been in a monogamous relationship of shorter duration and have a main partner with
positive views of condoms, with a lower likelihood of a prior pregnancy than those who did
not use condoms at baseline as seen in Table 1.
Contraceptive and condom use over time was dynamic with women experiencing dramatic
changes in both as represented in Figure 1. After initiation of hormonal methods at baseline,
overall condom use (condoms only or with a contraceptive method) dropped from 36% to
27% by 3 months. During the same period, dual method use increased from 5% to a peak of
20%. Over the 12 months, as women discontinued hormonal methods, there was a
substantial decrease in dual method use, an increase in condom only use, and little change in
overall condom use. Among condom users at baseline who discontinued condom use after
initiating an effective method and were no longer using that method at 1 year, 46% switched
back to condoms.

NIH-PA Author Manuscript

In the bivariate analysis of factors associated with dual method use at 12 months, being a
consistent condom user at baseline was associated with an odds ratio (OR) of 2.4 (95% CI:
1.6–3.6). Type of hormonal method chosen at baseline was not significantly associated with
dual method use nor was a prior pregnancy or history of an STI. Women who said their
partner thought condoms were “very important” or did not know how their partners felt
about condom use were more likely to be dual method users than women who said their
partner thought condoms were “not at all important” (OR 3.5, 95% CI: 1.9–6.3 and 2.5, 95%
CI: 1.3–4.8, respectively). In the multivariable model (adjusted for age, race/ethnicity, clinic
site), both baseline condom use and main partner’s views of condoms remained significant.
Those who thought it was “very likely” that they would get an STI in the next 3 months
were significantly less likely to be dual method users at 1 year (OR .5, 95% CI: .3–.99), as
seen in Table 2.
To determine whether those lost to follow-up may have affected our results, we conducted
sensitivity analyses that demonstrated that predictors of dual method use did not differ
significantly when those lost to follow-up were assumed not to be dual method users. When

J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

Goldstein et al.

Page 5

those lost to follow-up were considered dual users however, only condom use at baseline
remained a significant predictor of dual method use (data not shown).

NIH-PA Author Manuscript

Discussion

NIH-PA Author Manuscript

This study highlights the dynamic nature of the tradeoff between hormonal methods and
condom use in women initiating hormonal contraception as well as the influence of the male
partner on dual method use longitudinally. A tradeoff is a phenomenon whereby women
initially use condoms, begin a hormonal contraceptive, and subsequently discontinue
condom use. This was observed in the current study when, as women initiated a range of
effective hormonal contraceptives, dual method use increased; however, overall condom use
suffered as a result. This finding is consistent with the tradeoff between condom use and
hormonal methods demonstrated in a prior study [18]. The longitudinal nature of this study
allowed us to follow condom use over time as women initiated and later discontinued
hormonal contraception. By 12 months, the modest gains in dual method use were
diminished. What is important about this study is that we were able to demonstrate that not
only did women trade off condoms for hormonal methods, but as they discontinued the
hormonal methods, more than half (54%) failed to resume condom use, resulting in an
ultimate tradeoff of condoms for no method. This outcome is far from ideal. Given the
realities of the tradeoff between condoms and hormonal contraception, the use of condoms,
not only for STI protection but also as a backup method when hormonal contraception is
discontinued, should be underscored. Additionally, with the understanding that many
women discontinue these hormonal methods over time, promotion of long-acting reversible
contraception, including copper and progestin-releasing IUDs and implants, is essential.

NIH-PA Author Manuscript

Overall condom use decreased by nearly one third from baseline to 12 months, and although
hormonal method discontinuation was significant, there was a net gain in effective hormonal
contraceptive use. It should be noted that even on the most fertile day, the risk of a woman
becoming pregnant is less than half her risk of acquiring gonorrhea from an infected partner
[19]. At the same time, the risk of an STI only exists when a partner is infected, whereas the
risk of pregnancy (although varying in likelihood throughout a woman’s cycle) is present
with virtually all partners. Therefore, one might argue that it is possible for a subset of
women, those in committed monogamous relationships for instance, the tradeoff may be
justified. Unfortunately, we know from multiple studies that adolescents and young adults
often underestimate their risk of STIs [20,21]. Of note, having a high perceived risk of an
STI in the next 3 months was independently associated with reduced odds of being a dual
method user. Although it is counterintuitive, we speculate that other factors influence
contraceptive behaviors. In particular, gender-based power may explain why even though
women know they are at high risk, they have less agency in their relationships to negotiate
condom use and therefore have lower odds of dual method use [22].
Because of small sample size and the observational design of the study, we were unable to
compare STI acquisition rates among those that made the tradeoff compared with those that
did not. Although a recent analysis of a randomized intervention to increase dual method use
also failed to find a significant difference in biologic outcomes (STI and unintended
pregnancy incidence), the study found that those with the highest level of adherence had the
lowest incidence of STIs and unintended pregnancy [23]. Further studies are needed to
assess STI acquisition in the setting of initiation of hormonal methods.
Our results also highlight the strong influence that a woman’s main partner has on her
decision to be a dual method user, irrespective of her own views about dual method use.
This again may be related to the concept of relationship power imbalance and its impact on a
woman’s ability to negotiate condom use. Many of the associations observed in prior studies

J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

Goldstein et al.

Page 6

NIH-PA Author Manuscript

(age, African-American race, type of hormonal method initiated, partner concurrency, prior
pregnancy, and STIs) were not observed in our bivariate models. Partner’s attitude toward
condom use, in addition to condom use at baseline, were the only other significant predictors
after controlling for numerous other factors. This is consistent with findings from previous
research and highlights the importance of recognizing the role of the partner in contraceptive
choice [9,12,24]. Providers should inquire about partner attitudes about condom use and
work with women to develop techniques to address this component of the decision around
dual method use. Interestingly, women who reported they didn’t know their partner’s
attitudes about condoms were also more likely to report dual method use, suggesting that as
long as partners do not actively voice opposition to condoms, women are more likely to use
them along with another method. The need for the involvement of the partner in family
planning was emphasized in a recent study of adult women attending public family planning
clinics, which found that nearly two thirds of respondents were interested in some form of
partner involvement in their reproductive health planning [25]. It is clear from our findings
that the role of the partner is significant and that women do not make contraceptive
decisions in isolation. Providing couples-centered counseling may represent a way to
improve contraceptive, and more specifically, dual method use.

NIH-PA Author Manuscript

Our study has limitations that affect the interpretation of the data. Although our follow-up
rate was high at 88%, there were some differences in women who were lost to follow-up.
Women with follow-up data were more likely to be in school or working full time than those
who were lost to follow-up, indicating that our final study population may have been lower
risk. However, our sensitivity analysis demonstrated that even if none of those who were
lost to follow-up were dual method users, our results would be the same. The data were
obtained by self-report, which is susceptible to social desirability bias; however, the
majority of the data was collected by computer, which has been shown to improve reliability
for sensitive questions [26]. This study included a diverse group of women from urban and
suburban public family planning clinics in Northern California; it may not be generalizable
to other populations but this population represents an important demographic as they are at
high risk for experiencing unintended pregnancy and STIs.
Despite these limitations, this study provides a dynamic view of condom use among women
initiating hormonal methods and identifies key factors that could be addressed during family
planning visits to improve dual method use among women.

NIH-PA Author Manuscript

These results highlight a potential missed opportunity for family planning providers. With a
focus on getting women to initiate hormonal methods for pregnancy prevention, it is unclear
whether ample attention is given to method continuation and contingency planning in the
event of method discontinuation. Increasing dual method use is challenging as both
hormonal contraceptive use and condom use are complex behaviors with multiple mediating
factors that many women may have difficulty negotiating; that being said, it is crucial that
providers stress the importance of dual method use.

Acknowledgments
T.R.R. currently receives grant support from Teva Pharmaceutical Industries for research work on Plan B and Plan
B OneStep emergency contraception. This study was supported by the National Institute of Child Health and
Human Development grants R01 HD045480 and K24 HD057086-02. Additionally, this project was supported by
NIH/NCRR/OD UCSF-CTSI grant number TL1 RR024129. Its contents are solely the responsibility of the authors
and do not necessarily represent the official views of the National Institutes of Health. Results do not necessarily
represent the views of Planned Parenthood Federation of America, Inc. Findings from this study were reported in
poster format at the American College of Obstetricians and Gynecologists Annual Clinical Meeting in Washington,
D.C., and the National Predoctoral Clinical Research Training Program Meeting in St. Louis, MO in 2011. The
authors thank Cynthia Harper and Jody Steinauer for their advice on the analysis and reviewing early drafts of this
article.

J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

Goldstein et al.

Page 7

References
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript

1. Martinez GM, Copen CE, Abma JC. Teenagers in the United States: Sexual activity, contraceptive
use, and childbearing. 2006–2010 National Survey of Family. Growth Vital Health Stat. 2011;
23:1–44.
2. [Accessed April 16, 2011] The DCR report: Section A - Unplanned pregnancy in the United States
among all women. The National Campaign to Prevent Teen and Unplanned Pregnancy. Available
at: www.thenationalcampaign.org/resources/dcr/NATC_DCRreport.pdf
3. Finer, LB.; Zolna, MR. Unintended pregnancy in the United States: Incidence and disparities, 2006.
New York: Guttmacher Institute; 2011.
4. Finer LB. Unintended pregnancy among U.S. adolescents: Accounting for sexual activity. J Adolesc
Health. 2010; 47:312–4. [PubMed: 20708573]
5. Centers for Disease Control and Prevention. Sexually transmitted disease surveillance 2009.
Atlanta: CDC; 2010.
6. Mosher WD, Jones J. Use of contraception in the United States: 1982–2008. Vital Health Stat. 2010;
23:1–44.
7. Crosby RA, DiClemente RJ, Wingood GM, et al. Correlates of using dual methods for sexually
transmitted diseases and pregnancy prevention among high-risk African-American female teens. J
Adolesc Health. 2001; 28:410–4. [PubMed: 11336871]
8. Pazol K, Kramer MR, Hogue CJ. Condoms for dual protection: patterns of use with highly effective
contraceptive methods. Public Health Rep. 2010; 125:208–17. [PubMed: 20297747]
9. Riehman KS, Sly DF, Soler H, et al. Dual-method use among an ethnically diverse group of women
at risk of HIV infection. Fam Plann Perspect. 1998; 30:212–7. [PubMed: 9782043]
10. Roye CF. Condom use by Hispanic and African-American adolescent girls who use hormonal
contraception. J Adolesc Health. 1998; 23:205–11. [PubMed: 9763156]
11. Darney PD, Callegari LS, Swift A, et al. Condom practices of urban teens using Norplant
contraceptive implants, oral contraceptives, and condoms for contraception. Am J Obstet Gynecol.
1999; 180:929–37. [PubMed: 10203661]
12. Sangi-Haghpeykar H, Posner SF, Poindexter AN III. Consistency of condom use among lowincome hormonal contraceptive users. Perspect Sex Reprod Health. 2005; 37:184–91. [PubMed:
16380364]
13. Cushman LF, Romero D, Kalmuss D, et al. Condom use among women choosing long-term
hormonal contraception. Fam Plann Perspect. 1998; 30:240–3. [PubMed: 9782048]
14. Raine TR, Foster-Rosales A, Upadhyay UD, et al. One-year contraceptive continuation and
pregnancy in adolescent girls and women initiating hormonal contraceptives. Obstet Gynecol.
2011; 117:363–71. [PubMed: 21252751]
15. Minnis AM, van der Straten A, Gerdts C, Padian NS. A comparison of four condom-use measures
in predicting pregnancy, cervical STI and HIV incidence among Zimbabwean women. Sex Transm
Infect. 2010; 86:231–5. [PubMed: 19880972]
16. Lopez LM, Tolley EE, Grimes DA, Chen-Mok M. Theory-based interventions for contraception.
Cochrane Database Syst Rev. 2009
17. Rimer, BK.; Glanz, K. Theory at a Glance: A guide for Health Promotion Practice. Washington,
DC: National Institutes of Health, National Cancer Institute; 2005.
18. Ott MA, Adler NE, Millstein SG, et al. The trade-off between hormonal contraceptives and
condoms among adolescents. Perspect Sex Reprod Health. 2002; 34:6–14. [PubMed: 11990639]
19. Cates W Jr, Steiner MJ. Dual protection against unintended pregnancy and sexually transmitted
infections: what is the best contraceptive approach? Sex Transm Dis. 2002; 29:168–74. [PubMed:
11875378]
20. Ford CA, Jaccard J, Millstein SG, et al. Perceived risk of chlamydial and gonococcal infection
among sexually experienced young adults in the United States. Perspect Sex Reprod Health. 2004;
36:258–64. [PubMed: 15687084]
21. Ethier KA, Kershaw T, Niccolai L, et al. Adolescent women underestimate their susceptibility to
sexually transmitted infections. Sex Transm Infect. 2003; 79:408–11. [PubMed: 14573838]

J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

Goldstein et al.

Page 8

NIH-PA Author Manuscript

22. Tschann JM, Adler NE, Millstein SG, et al. Relative power between sexual partners and condom
use among adolescents. J Adolesc Health. 2002; 31:17–25. [PubMed: 12090961]
23. Peipert JF, Zhao Q, Meints L, et al. Adherence to dual method contraceptive use. Contraception.
2011; 84:252–8. [PubMed: 21843690]
24. Harvey SM, Henderson JT, Branch MR. Protecting against both pregnancy and disease: predictors
of dual method use among a sample of women. Women Health. 2004; 39:25–43. [PubMed:
15002881]
25. Zolna, MR.; Lindberg, LD.; Frost, JJ. Couple-focused services in publicly funded family planning
clinics: Identifying the need, 2009. New York: Guttmacher Institute; 2011.
26. Langhaug LF, Cheung YB, Pascoe SJ, et al. How you ask really matters: randomized comparison
of four sexual behaviour questionnaire delivery modes in Zimbabwean youth. Sex Transm Infect.
2011; 87:165–73. [PubMed: 20943824]

NIH-PA Author Manuscript
NIH-PA Author Manuscript
J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

Goldstein et al.

Page 9

IMPLICATIONS AND CONTRIBUTION

NIH-PA Author Manuscript

In our efforts to increase the use of hormonal contraceptives, condom use suffers and
contraceptive continuation is not optimal. This study highlights the large need for
effective interventions to improve long-term condom and contraceptive use among
adolescent and young adult women and their partners.

NIH-PA Author Manuscript
NIH-PA Author Manuscript
J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

Goldstein et al.

Page 10

NIH-PA Author Manuscript
NIH-PA Author Manuscript

Figure 1.

Condom use over time with the initiation of hormonal methods.
Hormonal methods
only.a
Overall condom use.b
Dual method use.
Condom only. aAlso
includes small percentage of women (1.8%) using other effective methods (intrauterine
devices, implants). bIncludes condom only and dual method use.

NIH-PA Author Manuscript
J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

NIH-PA Author Manuscript
391
390

18–19

20–24

320
407
141
193

Latina

African American/black

Asian/Pacific Islander

Multiracial/other

198

Not working or in school

J Adolesc Health. Author manuscript; available in PMC 2013 August 16.
139
382

Vallejo

Richmond

82
56
12
29

Pill

Patch

Ring

DMPA

Method used at last sex (n = 1,190)***

Reproductive history

655

312

Hayward

Low income neighborhood (n = 1,171)a

361

Oakland

Clinic site (n = 1,194)

996

In school or employed (full or part time)

School and employment status (n = 1,194)

133

White

Race/ethnicity (n = 1,194)

413

1,194

Total

15–17

Age (years)**

Sociodemographics

Characteristic

2.4

1.0

4.7

6.9

55.9

11.6

26.1

30.2

16.6

83.4

16.2

11.8

34.1

26.8

11.1

32.7

32.8

34.6

244

138

120

125

371

150

105

116

146

170

432

2.3

3.5

7.0

58.1

31.9

11.3

27.8

28.9

14.1

85.9

17.4

10.0

34.7

24.3

13.7

26.9

33.8

39.4

36.2

Condom user

411

244

192

236

137

625

118

257

215

274

245

243

762

2.5

1.4

5.4

6.9

54.7

32.0

11.8

25.2

31.0

18.0

82.0

15.5

12.9

33.7

28.2

9.7

36.0

32.2

31.9

63.8

Non-condom user

NIH-PA Author Manuscript

Characteristics of participants by condom use status at baseline

NIH-PA Author Manuscript

Table 1
Goldstein et al.
Page 11

NIH-PA Author Manuscript
56

Dual method users

140
352
184

7–12
>12
No sex yet

NIH-PA Author Manuscript

J Adolesc Health. Author manuscript; available in PMC 2013 August 16.
12
255
4

STI/AIDS prevention
Both
Other

307
327

Very important
Somewhat important

Main partner views of condoms (n = 1,008)d,***

188

Pregnancy prevention

Reason for condom use at last sex (n = 459)c,**

Attitudes toward condom use

518

0–6

Time had sex with main partner (months) (n = 1,194)***

170

319

Very likely
Partner concurrency (sex with man other than main partner) (n = 1,194)*

875

Not at all/a little likely/somewhat

Perceived STI risk in next 3 months with main partner (n = 1,194)

255

257

DMPA

Prior STI (n = 1,169)

232

Ring

586

339

Patch

Prior pregnancy (n = 1,194)***

366

Pill

Baseline method initiated (n = 1,194)*

993

1,194

None/withdrawal

Other effective methodb

Total

32.4

30.5

55.6

2.6

15.4

29.5

11.7

43.4

14.2

26.7

73.3

21.8

49.1

21.5

19.4

28.4

30.7

4.7

83.4

115

204

211

133

228

118

314

165

112

158

366

432

31.6

56.0

59.8

2.3

37.7

15.3

22.5

9.5

52.8

10.9

27.3

72.7

19.7

38.2

20.4

17.1

25.9

36.6

84.9

36.2

Condom user

212

103

118

255

290

123

201

561

172

421

169

158

227

208

627

762

32.9

16.0

2.8

41.5

3.8

51.9

15.5

33.5

13.0

38.1

16.1

26.4

73.6

23.0

55.2

22.2

20.7

29.8

27.3

82.6

63.8

Non-condom user

NIH-PA Author Manuscript

Characteristic

Goldstein et al.
Page 12

NIH-PA Author Manuscript
212
119
20

Neither
Disagree
Strongly disagree

1.7

17.9

35.1

35.3

5.3

31.8

132

185

432

2.1

7.7

16.3

30.8

43.1

11.8

36.2

Condom user

142

285

234

278

762

1.5

11.3

18.7

37.6

30.9

7.9

43.2

63.8

Non-condom user

Asked only of those with a main partner.

Asked of those who used a condom at last sex.

Includes one woman who reported using a once-monthly injectable contraceptive.

Defined as living in a zip code where the proportion of families living below the federal poverty level is greater than the national average.

p < .001.

p < .01,

***

p < .05,

DMPA = depot medroxyprogesterone; STI = sexually transmitted infection.

417

Agree

Strongly agree

419

Don’t know
Believes should always use condoms, even if on birth control (n = 1,187)***

321

Not at all important

1,194

Total

NIH-PA Author Manuscript

Characteristic

Goldstein et al.
Page 13

NIH-PA Author Manuscript

J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

Goldstein et al.

Page 14

Table 2

Predictors of dual method use at 12 months

NIH-PA Author Manuscript

Unadjusted OR (95% CI)a
N = 1,018

Adjusted OR (95% CI)b
N = 1,012

<80% of time

1 (reference)

≥80% of time

2.40 (1.62–3.58)***

2.01 (1.28–3.14)**

Prior sexually transmitted infectionc

1.35 (.93–1.96)

1.55 (.94–2.55)

Prior pregnancy

1.11 (.75–1.65)

1.18 (.73–1.88)

Not at all/a little/somewhat likely

1 (reference)

Very likely

.89 (.56–1.42)

.54 (.30–.99)*

.77 (.42–1.41)

.73 (.38–1.40)

Variable
Baseline condom use

Perceived STI risk in next 3 months

Partner concurrency (other than main)
Main partner views of condoms

NIH-PA Author Manuscript

Not at all important

1 (reference)

Somewhat important

1.65 (.86–3.20)

1.55 (.78–3.10)

3.45 (1.88–6.34)***

2.89 (1.47–5.71)**

2.45 (1.25–4.81)*

2.99 (1.35–6.64)**

1 (reference)

Very important
Don’t know
Believes should always use condoms, even if on birth controld
Strongly disagree
Disagree

.29 (0.05–1.72)

.48 (.07–3.13)

Neither

.91 (019–4.27)

1.54 (.30–7.78)

Agree

.93 (.20–4.24)

1.41 (.29–6.89)

Strongly agree

.97 (.21–4.41)

1.07 (.22–5.15)

p < .05,

p < .01,

***

p < .001.

n = 1,018 as 176 women were missing data for dual method use at 12 months.

Adjusted for age, race/ethnicity, and clinic site.

NIH-PA Author Manuscript

Individuals who were missing on STI history were categorized as missing as to include in multivariable analysis.

n = 1,012 because of missing data.

J Adolesc Health. Author manuscript; available in PMC 2013 August 16.

Journal of Adolescent Health 39 (2006) 388 –395

Original article

The Developmental Association of Relationship Quality, Hormonal
Contraceptive Choice and Condom Non-Use among Adolescent Women
M. Aaron Sayegh, Ph.D., M.P.H.a,b,* J. Dennis Fortenberry, M.D., M.S.b,
Marcia Shew, M.D., M.P.H.b, and Donald P. Orr, M.D.b
a

Organizational Wellness and Learning Systems, Fort Worth, Texas
b
Indiana University School of Medicine, Indianapolis, Indiana
Manuscript received June 27, 2005; manuscript accepted December 16, 2005

Abstract

Purpose: Consistent condom use is critical to efforts to prevent sexually transmitted infections
among adolescents, but condom use may decline as relationships and contraceptive needs change.
The purpose of this research is to assess changes in condom non-use longitudinally in the context
of changes in relationship quality, coital frequency and hormonal contraceptive choice.
Methods: Participants were women (aged 14 –17 years at enrollment) recruited from three urban
adolescent medicine clinics. Data were collected at three-month intervals using a face-to-face
structured interview. Participants were able to contribute up to 10 interviews, but on average
contributed 4.2 interviews over the 27-month period. Independent variables assessed partnerspecific relationship quality (five items; scale range 5–25; ␣ ⫽ .92, e.g., this partner is a very
important person to me); and, number of coital events with a specific partner. Additional items
assessed experience with oral contraceptive pills (OCP) use and injected depo medroxy-progesterone acetate (DMPA). The outcome variable was number of coital events without condom use during
the past three months. Analyses were conducted as a three-level hierarchical linear growth curve
model using HLM 6. The Level 1 predictor was time, to test the hypothesis that condom non-use
increases over time. Level 2 predictors assessed relationship quality and coital frequency across all
partners to assess hypotheses that participants’ condom non-use increases over time as a function of
relationship quality and coital frequency. Level 3 predictors assessed the participant-level influence
of OCP or DMPA experience on time-related changes in condom non-use.
Results: A total of 176 women reported 279 sex partners and contributed 478 visits. Both average
coital frequency and average condom non-use linearly increased during the 27-month follow-up. At
any given follow-up, about 35% reported recent OCP use, and 65% reported DMPA use. HLM
analyses showed that condom non-use increased as a function of time (␤ ⫽ .12; p ⫽ .03, Level 1
analysis). Increased condom non-use over time was primarily a function of increased coital
frequency (␤ ⫽ .01; p ⫽ .00), although higher levels of relationship quality were associated with
increased condom non-use at enrollment (␤ ⫽ .44; p ⫽ .00, Level 2 analysis). The temporal rise in
condom non-use significantly increased among DMPA users (␤ ⫽ .06; p ⫽ .00) but not OCP users
(Level 3 analysis) (␤ ⫽ ⫺.04; p ⫽ .06).
Conclusions: Developmentally, relationship characteristics and coital frequency appear to have increasing weight in decisions about condom use. Hormonal contraceptive methods are not equivalently
associated with the overall temporal decline in condom use. Future research associated with dual
contraceptive/condom use should address differential factors associated condom use in combination with
different hormonal methods. © 2006 Society for Adolescent Medicine. All rights reserved.

Keywords:

Hormonal contraception; Condom use; Relationship quality

*Address correspondence to: Dr. M. Aaron Sayegh, Organizational Wellness and Learning Systems, 4413 Overton Terrace,

Fort Worth, TX 76109.
E-mail address: [email protected]

M.A. Sayegh et al. / Journal of Adolescent Health 39 (2006) 388 –395

The prophylactic functions of condoms include prevention of sexually transmitted infections (STI) as well as
conception [1]. This dual function makes condoms unique
among the contraceptive methods available to adolescents.
However, many adolescents, especially women, prefer more
reliable, coitus-independent contraceptive methods that do
not reduce STI risk [2]. In fact, hormonal contraceptive
methods may increase risk of some STI, making condom
use even more important [3]. Thus, so-called “dual use” of
coitus-independent contraception plus condom use with
each coitus has become a public health standard [4]. This
ideal has been difficult to achieve, however, and fewer than
25% of adolescent women consider themselves to be “dual
users” [5,6]. Many adolescent women view dual use as a
“trade-off” between intimacy and decreased perceived STI
risk. These women are especially unlikely to use condoms
in addition to hormonal contraceptive methods [7].
Issues surrounding pregnancy and STI prevention become additionally complex because choices about contraception and condom use are not static characteristics of
adolescents or their sexual relationships. Condom use is
more common with a new sexual partner, and during the
early weeks of a relationship [8,9]. Adolescents discontinue
condom use quickly, perhaps less than one month, suggesting that consistent condom use is a relatively short-lived
characteristic of many sexual relationships [10]. Among
adolescent women, condom use is less likely in partnerships
characterized as relatively higher in emotional affiliations
[11,12]. Although coital frequency is typically higher in
more stable relationships, greater coital frequency appears
to be associated independently with increased levels of
non-condom use, even when relationship characteristics are
controlled [12,13].
A final factor that may affect condom use is the specific
type of coitus-independent contraceptive method. The most
widely used methods— oral contraceptive pills (OCP) and
depot medroxy-progesterone acetate (DMPA)–are similar in
terms of contraceptive effectiveness but differ markedly in
terms of method characteristics and user perceptions. For
example, up to one-half of OCP users miss enough doses to
place them at risk for pregnancy, with the implied need for
a back-up contraceptive method [14]. Thus, the demand of
daily pill-taking (and awareness of failures in pill-taking)
mandates more attention to contraception and STI prevention issues while the certainty of contraceptive protection
and prolonged intervals between DMPA injections may
diminish such attention [15]. Therefore, accumulation of
experience with a method such as DMPA may be associated
with increasing levels of condom non-use, whereas OCP
experience would be less likely to change levels of condom
non-use.
Existing research lacks an understanding of the effects of
developmental change in condom use and contraceptive
behaviors. The average age of first sexual intercourse for
American women is about age 16, thus a substantial portion

389

of middle and late adolescent development may be accompanied by sexual activity [16]. By age 20, an average of five
lifetime sex partners is reported. These partnerships are
usually sequential, allowing for an accumulation of experience with romantic and sexual relationships, changing personal and interpersonal motivations for sex, and changes in
motivations for contraception [17]. Condom use behaviors
appear to decline over time, not only within a given relationship, but in each succeeding relationship [9,10].
These developmental contexts of hormonal contraception and condom use warrant a more detailed understanding
of their development over time. The purpose of this research, then, is to evaluate the short-term (within threemonth intervals) and long term (over 27 months) changes in
condom non-use in the context of partner-specific relationship quality, partner-specific coital frequency, and hormonal
contraceptive choice. A latent growth curve (LGC) approach using hierarchical linear modeling (HLM) is used to
allow the examination of patterns of change (e.g., linear or
non-linear) as well as testing of hypotheses about potential
predictors of these changes. LGC allows individual and
group change to be modeled by using a varying number and
spacing of data points across time, which is indicative of
these data [18]. This method also allows the information
from all the sex partners during any time period to be
incorporated and analyzed in a predictive model.
Methods
Study design and procedures
Data were collected as part of a larger longitudinal study
of risk and protective factors (initiated in 1999) associated
with sexually transmitted infections among women in middle adolescence. Briefly, the larger study consisted of an
enrollment visit and follow-up clinic visits each three
months during a 27-month study period (up to 10 visits
total). At each visit, a structured face-to-face interview with
trained research assistants provided detailed information
regarding sexual and contraceptive behaviors in the previous three-month period. Informed consent was obtained
from each participant and permission obtained from a parent
or legal guardian. This research was approved by the institutional review board of Indiana University/Purdue University at Indianapolis – Clarian.
Participants were adolescent women receiving health
care in three primary health clinics in Indianapolis. These
clinics serve mostly African-American neighborhoods of
lower- and middle-income residents in areas with high rates
of teen pregnancy and sexually transmitted infections. For
example, the proportion of African-Americans in census
tracts served by participating clinics was 78%, and the
median household income was $28,000. The 2004 Chlamydia rates for zip codes in which the participating clinics
are located ranged from 469/100,000 to 1656/100,000. A

390

M.A. Sayegh et al. / Journal of Adolescent Health 39 (2006) 388 –395

majority (⬎ 75%) of female clinic patients are AfricanAmerican, and a large majority of clinic patients receive
some form of public assistance for health care (⬎ 85%).
Study participants resembled the racial-ethnic composition
of participating clinics in that 87% of the sample reported
race as African-American.
Clinic patients were eligible for study participation if
they were aged between 14 and 17 years at enrollment,
spoke English, and were not pregnant at enrollment. However, participants who became pregnant were continued in
the study. The age range of 14 –17 years was chosen because of high rates of initiation of sexual activity. Thus,
lifetime sexual experience was not an enrollment requirement because many initially sexually inexperienced could
be expected to become sexually active during the follow-up
period. For this analysis, women who became pregnant or
who were not using a hormonal contraceptive method were
excluded.
Measures
The primary outcome measure was condom non-use,
defined as the total number of coital events (during the
previous three months) unprotected by condoms. We chose
this measure as a reflection of potential exposure to STI that
does not confound levels of condom non-use with levels of
coital frequency [19].
Independent variables consisted of partner-specific relationship quality and partner-specific coital frequency and
partner-independent measures of hormonal contraceptive
choice. At enrollment and each follow-up visit, participants
were asked to identify sex partners by first name or initial,
last name, nickname, any contact information, and street
address if known. This information was used to create
unique partner identifiers in order to link relationship-specific attitudes and behaviors to their specific relationships.
Partner-specific variables included relationship quality
and coital frequency. Relationship quality assesses positive
emotional and affiliational aspects of a relationship. The
additive index consists of five items coded as “strongly
disagree,” “disagree,” “agree” or “strongly agree,” (scale
range 5–25; ␣ ⫽ .92) with higher scores indicative of
greater relationship quality. Example items include [this
partner] is a very important person to me and I enjoy
spending time with [this partner].
Coital frequency reflected the total number of coital
events with a given partner. Coital frequency was assessed
by asking “How many times in the past three months did
you have sex?” Responses were recorded verbatim for each
partner. Participants were asked for an approximate number,
and “missing” was entered when a precise estimate was not
provided. Information on all the sex partners in the past
three months were used for this analysis in order to assess
the total effects of relationship quality and coital frequency
on the outcome measure.

The individual’s experience with OCP or DMPA was
assessed as the cumulative number of visits in which OCP
or DMPA was reported as the method of contraception used
in the previous three months. Values range from 0 to 10
with lower values representing less DMPA or OCP experience and higher values representing greater OCP or DMPA
experience.
Statistical methods
Analyses were conducted as a three-level hierarchical
linear growth curve (LGC) model using HLM 6 [20]. LGC
analysis allows examination of patterns of change with
repeated observations of nested variables and with a varying
number of data points across time. In the current application, LGC is used to estimate individual growth trajectories
of condom non-use by fitting a regression line to individual
observations of condom non-use over time. The information
from all the individual curves is used to create a summary
measure of condom non-use. This summary measure is what
is referred to as the “latent growth curve.” The curve is latent
because it captures group level growth based on the relationships between the observed measures across time [18].
The HLM approach to LGC conceptualizes time as
nested within the individual. LGC with HLM allows examination of the influence of specific variables on these curves.
Present analyses consisted of creating a taxonomy of nested
models as suggested by other researchers employing the
methodology [18]. These models are created by sequential
addition of predictors, with examination of statistical significance of the predictors and changes in overall model fit.
Improvements in fit were assessed by examination of the
changes in pseudo-R2, the deviance statistic, Akaike Information Criterion (AIC), and the Bayesian Information Criterion (BIC) [18]. The pseudo-R2 represents changes in the
percentage of unexplained variance as model components
are added. The AIC and BIC assess model fit as a function
of model complexity, with progressively smaller values for
deviance, AIC and for BIC indicating better fit.
LGC produces two latent factors (the intercept and the
slope), which together form the trajectory. The intercept is
the start, or initial, value. The slope represents the rate of
change, or growth, over time. A positive sign for the slope
indicates an increase over time. A negative sign indicates a
decrease over time. HLM assesses the influence of factors at
different levels in the model on the intercept and slope.
Here, condom non-use is assessed over time (measured by
the sequence of three-month visits) and is specified as
nested within partner-specific variables (i.e., relationship
quality and coital frequency) because individual participants
may have more than one partner during any given threemonth period. Partner-specific variables are specified as
nested within the duration of experience with a contraceptive method choice because a given method applies equally
to all partners. Thus (in the language of LGC), the Level 1

M.A. Sayegh et al. / Journal of Adolescent Health 39 (2006) 388 –395
Table 1
Descriptive statistics for Level 1, Level 2 and Level 3
Variable name

Mean

Visit-specific variables
(Level 1; n ⫽ 478
visits)
Condom nonuse
Time
Partner-specific variables
(Level 2; n ⫽ 275
partners)
Relationship quality
Coital frequency
Participant-specific
variables (Level 3; n
⫽ 175 participants)
Hormonal contraceptive
experience
DMPA
OCP

Minimum

Maximum

6.95
—

14.38
—

0
0 months
(enrollment)

60.00
27 months

18.53
7.57

4.22
10.86

6.00
0

25.00
60.00

2.33
1.24

2.59
1.67

0
0

10.00
10.00

391

where ␥0ij is the condom non-use for adolescent j with
partner i at time zero; ð1ij is the condom non-use rate of
change over the length of the study; ␥10j is the mean rate of
change across each partner for adolescent j; r1ij and r2ij are
random errors (associated with the intercept and subsequent
slope).
Three Level 2 models were evaluated: partner-specific
coital frequency only; partner-specific relationship quality
only; and simultaneous inclusion of both coital frequency
and relationship quality.
Level 3 captures the variability between adolescents. In
this case, Level 3 captures the variability between the
groups (DMPA or OCP) to which the adolescents belong.
The general conditional Level 3 model with fixed effects
and no covariates:

␥00j ⫽ ␥000 ⫹ u00j

Note: Sample represents 176 participants (Level 3 variables) with 278
sexual partners (Level 2 variables) and 478 visits (Level 1 variables).
* 4.10 per partner.

predictor was time, to test the hypothesis that condom nonuse increases over time (i.e., during each three-month period
over the 27 months). In the taxonomy of models, these
analyses represent temporal change in condom non-use
without assessment of influence by other predictors (see
Time Only model, Table 2).
Level 1 generates individual change trajectories (intercept and slope) for each outcome of interest (here condom
non-use). At Level 1 an individual growth model of condom
non-use at time t with partner i for participant j is in the
form:
Y 1ij ⬅ ␥0ij ⫹ ␥1ij共Time兲tij ⫹ etij
where Ytij is the condom non-use at time t with partner i for
adolescent j; ␥0ij represents the expected level of condom
non-use for adolescent j at Time zero (i.e., the initial status
or intercept); ␥1i is the rate of change in the condom non-use
during the three-month interval with partner i for adolescent
j; (Time)tij is 0, 1, 2, 3, 4, . ѧ12; ␥2ij is the rate of change
with partner i for adolescent j; and, etij represents the withinperson residual.
Level 2 predictors were coital frequency (with each partner)
and relationship quality (with each partner) to assess hypotheses that participants’ condom non-use increases over time as
a function of partner-specific coital frequency and partnerspecific relationship quality. Level 2 generates partner level
change trajectories (mean intercept, intercept variance, mean
slope, slope variance). The general unconditional Level 2
model with fixed effects and no covariates is:

␥0ij ⫽ ␥00j ⫹ r0ij
␥1ij ⫽ ␥10j ⫹ r1ij

␥10j ⫽ ␥100 ⫹ r10j
where ␥00j represents the mean initial status within adolescent j; ␥000 is the overall mean initial status; ␥10j is the mean
rate of change within adolescent j; ␥100 is the overall mean
growing rate; and u00j and u10j are Level 3 random error
terms (associated with the intercept and slope). As can be
seen, the Level 2 slopes and intercepts become the outcomes
of the Level 3 model. Level 3 predictors assessed the additional influence of OCP or DMPA experience on timerelated changes in condom non-use.
Results
The original sample contained 237 women providing 732
enrollment and visits at three-month intervals over the 27month follow-up. In order to focus appropriately on the
effects of hormonal contraceptive choice, and partner-specific relationship quality and partner-specific coital frequency on change in condom non-use over time, analyses
were limited to the 176 women using OCP or DMPA (Level
3 variables), during relationships with 279 sex partners (and
thus 275 assessments of partner-specific relationship quality
and partner-specific coital frequency; Level 2 variables)
with 478 visits. On average, participants contributed 4.2
visits each, with a range from 2 to 10.
Overall, the average level of condom non-use was 4.10
unprotected coital events per partner during any given threemonth period. The average number of coital events per
partner was 7.57 during any given three-month period. Average relationship quality was 18.5 per partner during any
given three-month period. At each visit, DMPA use was
more common than that of OCP (Table 1).
Average condom non-use and average coital frequency
increased linearly over the 27-month follow-up period (Figure 1). For example, the average number of unprotected
coital events per partner was 6.13 at enrollment and 9.54 by
the end of the 27-month follow-up period. Average rela-

392

M.A. Sayegh et al. / Journal of Adolescent Health 39 (2006) 388 –395

Table 2
Taxonomy of models— condom nonuse over time

Intercept
Time
Relationship quality
Coital frequency
DMPA experience
OCP experience
Pseudo R-square

Parametera

Time
only
model

Relationship
quality
model

Coital frequency
model

Coital
frequency-relationship
quality model

DMPA
model

OCP model

y00
y10
y01
y11
y101
y101

3.64 (.82)
.2 (.09)
—
—
—
—
.07
.09
—
—
—
3761.5
3715.9
3728.3

3.52 (.82)
.18 (.08)
.65 (.17)
—
—
—

4.14 (.94)
.13* (.07)
—
.02 (.00)
—
—

3.78 (.81)
.16 (.06)
.44 (.15)
.01 (.00)

3.83 (.81)
.12 (.05)
.44 (.15)
.01 (.00)
.06 (.02)

3.63 (.78)
.2 (.06)
.44 (.15)
.01 (.00)
—
⫺.04* (.02)

—
—
—
—
Deviance
AIC
BIC

—
—

.11
—
—
3747.8
3693.1
3699.0

.13
—
3716.8.9
3704.9
3706.3

.15
3708.4
3701.9
3658.8

.15
3705.5
3653.7
3656.3

3711.2
3659.6
3658.2

Note. Numbers are beta and (standard errors). All beta coefficients are statistically significant at p ⬍ .05 unless otherwise noted.
See Appendix for guide to symbols.
b
Final model includes time (Level 1), coital frequency and relationship quality (Level 2) plus contraceptive method experience (Level 3).
* p ⫽ .06.
a

tionship quality remained relatively stable during the follow-up period (Figure 2).
Analyses showed that condom non-use increased as a
function of time (␤ ⫽ .12; p ⫽ .03, Level 1 analysis). A
convention of LGC analysis is to report parameter estimates
as unstandardized betas [16]. Estimates should be understood in reference to their original metrics: for example, a

one-unit increase in time is associated with a .12-unit increase in condom non-use.
Both coital frequency and relationship quality separately
resulted in improved model fit compared to the Time Only
model. Addition of both coital frequency and relationship
quality yielded improved fit over models with only coital
frequency or relationship quality (Table 2). The source of

Figure 1. Partner-specific coital frequency and condom non-use over the 27 months.

M.A. Sayegh et al. / Journal of Adolescent Health 39 (2006) 388 –395

393

Figure 2. Mean partner-specific relationship quality over 27 months.

influence differed, however. Increased condom non-use
over time was a function of increased coital frequency (␤ ⫽
.01; p ⫽ .00), while higher levels of relationship quality
were associated with increased condom non-use at enrollment (␤ ⫽ .44; p ⫽ .00, Level 2 analysis).
Addition of measures of OCP and DMPA experience
produced significant increases in model fit. The pseudo-R2
estimates showed substantial reductions in model variance
with the addition of model components (Table 2). The AIC
and BIC of the final model decreased (compared to the
models containing only coital frequency and relationship
quality [Table 2]), indicating improved model fit.
Increased OCP experience (compared to DMPA) was
associated with decreased rate of change in condom non-use
(␤ ⫽ ⫺.04; p ⫽ .06) (Table 2). Increased DMPA experience
was associated with increased rate of change in condom
non-use (␤ ⫽ .06; p ⬍ .00). The contrast in latent growth
curves for OCP users versus DMPA users is illustrated in
Figure 3. This figure shows a much steeper rate of increase
in condom non-use for participants who reported more cumulative experience with DMPA, compared to those with
greater OCP cumulative experience.

exposure. Hormonal contraceptive methods are not equivalently associated with the overall temporal decline in condom use: more experienced DMPA users become substantially less likely to use condoms over time, while
experienced OCP users maintain relatively stable rates of
condom use.
The finding that condom non-use becomes more frequent
over time is consistent with other research. For example,
condom use declines not only within relationships, but becomes lower with succeeding relationships [9]. In other
research, the time required for levels of condom use in new
relationships to approximate that of established relationships is about three weeks [10].

Discussion
The results of this study clearly demonstrate a temporal
increase in condom non-use during a 27-month follow-up.
Both relationship characteristics and coital frequency influence the rate of increase in condom non-use, with adolescent women perceiving higher relationship quality and reporting greater coital frequency at greater risk for STI

Figure 3. Growth curves of condom non-use for OCP coverage, and DMPA
coverage 27 months.

394

M.A. Sayegh et al. / Journal of Adolescent Health 39 (2006) 388 –395

Developmentally, adolescent romantic relationships appear to shift from a primarily self-focus during early adolescence to a substantially couples-perspective during middle and late adolescence [21]. Parental control and
supervision is likely stricter for younger than for older
adolescents. The less relationship-oriented sexual activity
and tighter parental control of early and middle adolescence
makes condoms an expedient but reasonably effective form
of contraception. From a relationship perspective, an increasing sense of oneself as a member of a couple rather
than an individual in a pair means that factors such as trust,
intimacy, and relationship maintenance become important
functions of sex within relationships [22]. Even if adolescents accurately estimated STI and pregnancy risk in sexual
relationships, condom use works against the romantic ideals
that are part of many adolescent sexual relationships [23].
From a pragmatic perspective, condom non-use may increase over time simply as a function of interest in and
access to effective methods that separate sex from pregnancy and STI prevention [24].
These findings are also consistent with others in that both
relationship characteristics and coital frequency are important influences on levels of condom use within a relationship
[11,12,19]. One important contribution of the current research is evidence that coital frequency and relationship
quality have different effects on levels of condom non-use.
Relationship quality appears to affect condom non-use by
its influence on initial levels of non-use, so that higher levels
of relationship quality were associated with more frequent
condom non-use at study entry. Coital frequency, on the
other hand, influenced the rate of change of condom non-use
over time. This suggests that relationship characteristics
present early in a relationship, rather than relationship duration, are important early influences on a couple’s condom
use behaviors. As shown elsewhere, relationship quality is
an important correlate of coital frequency, but coital frequency has the most direct proximal effect on condom
non-use [25].
The data also demonstrate important differences in the
association of contraceptive method choice and condom
non-use, with increased non-use primarily a characteristic of
participants with more experience with DMPA. An earlier
study found low levels of condom use among users of
contraceptive implants [26]. However, that study did not
assess condom use among users of other contraceptive
methods. Reasons for the differential effect of contraceptive
method on condom non-use are not known, although the
marked difference in demand characteristics of the methods
is a plausible explanation [15]. Differential effects suggest
that “dual contraceptive method use” is not simply an issue
of a combination of any coitus-independent method with
condom use. Rather, clinical and public health efforts to
encourage dual method use may need to include evaluation
of method choice as well as relationship and sexual behavior factors. Contraceptive patches or vaginal contraceptive

rings are coitus-independent but require weekly attention
(and do not typically affect menses), future research should
address condom non-use among users of these methods.
Limitations
The sample is relatively homogenous in terms of geography (drawn from residents of a single urban area), race
(mostly African-American) and socioeconomic status
(mostly lower and middle socioeconomic status). Although
the sample reasonably resembles the clinical population
from which it was assembled, generalization of results
should be made with caution. Secondly, a limited number of
predictors of condom non-use were examined. Condom use
has been linked to a host of socioeconomic, cultural, familial, psychological, and behavioral factors [27]. However,
few of these factors have been investigated in the context of
complex longitudinal models. We chose to focus on factors
likely to have proximal influence on condom use behaviors.
Thirdly, a limited number of contraceptive methods were
assessed. If other coitus-independent methods become increasingly used with condoms these methods would be of
interest. Finally, the resulting growth curves refer to grouplevel change over time in condom non-use. These results
cannot be interpreted to represent developmental trajectories of individual participants.
Conclusion
The sometimes competing needs for effective contraception and effective STI prevention represent complex behavioral targets to achieve even for a short period of time.
These data demonstrate, however, that condom use, especially, represents a developmental “moving target” subject
to change over time and in response to changes in relationships, sexual behaviors and contraceptive choices. Perhaps
the most important message to derive from our data is the
suggestion that efforts to encourage condom use must be
persistent, and must be adjusted to the relational and contraceptive needs of a given time.
References
[1] Holmes KK, Levine R, Weaver M. Effectiveness of condoms in
preventing sexually transmitted infections. Bull World Health Organ
2004;82:454 – 61.
[2] Grady WR, Klepinger DH, Nelson-Wally A. Contraceptive characteristics: the perceptions and priorities of men and women. Fam Plann
Perspect 1999;31:168 –75.
[3] Baeten JM, Nyange PM, Richardson BA, et al. Hormonal contraception and risk of sexually transmitted disease acquisition: results from
a prospective study. Am J Obstet Gynecol 2001;185:380 –5.
[4] Bearinger LH, Resnick MD. Dual method use in adolescents: a
review and framework for research on use of STD and pregnancy
protection. J Adolesc Health 2003;32:340 –9.
[5] Santelli JS, Warren CW, Lowry R, et al. The use of condoms with
other contraceptive methods among young men and women. Fam
Plann Perspect 1997;29:261–7.

M.A. Sayegh et al. / Journal of Adolescent Health 39 (2006) 388 –395
[6] Harvey SM, Henderson JT, Branch MR. Protecting against both
pregnancy and disease: predictors of dual method use among a sample of women. Women Health 2004;39:25– 43.
[7] Ott MA, Adler NE, Millstein SG, et al. The trade-off between hormonal contraceptives and condoms among adolescents. Perspect Sex
Reprod Health 2002;34:6 –14.
[8] Manning WD, Longmore MA, Giordano PC. The relationship context
of contraceptive use at first intercourse. Fam Plann Perspect 2000;
32:104 –10.
[9] Ku L, Sonenstein FL, Pleck JH. The dynamics of young men’s
condom use during and across relationships. Fam Plann Perspect
1994;26:246 –51.
[10] Fortenberry JD, Tu W, Harezlak J, et al. Condom use as a function of
time in new and established adolescent sexual relationships. Am J
Public Health 2002;92:211–3.
[11] Plichta SB, Weisman CS, Nathanson CA, et al. Partner-specific condom use among adolescent women clients of a family planning clinic.
J Adolesc Health 1992;3:506 –11.
[12] Katz BP, Fortenberry JD, Zimet GD, et al. Partner-specific relationship characteristics and condom use among young people with sexually transmitted infections. J Sex Res 2000;37:69 –75.
[13] Williams M, Ross MW, Bowen AM, et al. An investigation of
condom use by frequency of sex. Sex Transm Infect 2001;77:433–5.
[14] Potter L, Oakley D, de Leon-Wong E, Cañamar R. Measuring compliance among oral contraceptive users. Fam Plann Perspect 1996;
28:154 – 8.
[15] Breheny MMA, Stephens C. Barriers to effective contraception and
strategies for overcoming them among adolescent mothers. Public
Health Nurs 2004;21:220 –7.
[16] Abma JC, Sonenstein FL. Sexual activity and contraceptive practices
among teenagers in the United States, 1988 and 1995. Vital Health
Stat 2001;23:1–79.

395

[17] Howard MM, Fortenberry JD, Blythe MJ, et al. Patterns of sexual
partnerships among adolescent females. J Adolesc Health 1999;24:
300 –3.
[18] Singer JD, Willett JB. Applied Longitudinal Data Analysis: Modeling
Change and Event Occurrence. New York, NY: Oxford University
Press, 2004.
[19] Crosby R, DiClemente RJ, Holtgrave DR, Wingood GM. Design,
measurement, and analytical considerations for testing hypotheses
relative to condom effectiveness against non-viral STIs. Sex Transm
Infect 2002;78:228 –31.
[20] Raudenbush SW, Bryk AS, Cheong YF, Congdon R. HLM 6: Hierarchical Linear and NonLinear Modeling. Chicago, IL: Scientific
Software International, 2004.
[21] Carver K, Joyner K, Udry JR. National estimates of adolescent
romantic relationships. In: Florsheim P, ed. Adolescent Romantic
Relations and Sexual Behavior: Theory, Research, and Practical Implications. Mahwah, NJ: Lawrence Erlbaum Associates, 2003:56 – 89.
[22] Ott MA, Adler NE, Millstein SG, et al. The trade-off between hormonal contraceptives and condoms among adolescents. Perspect Sex
Reprod Health 2002;34(7):6 –14.
[23] Tolman DL. Object lessons: romance, violation and female adolescent sexual desire. J Sex Educ Ther 2000;25:70 –9.
[24] Weisman CS, Plichta S, Nathanson CA, et al. Consistency of condom
use for disease prevention among adolescent users of oral contraceptives. Fam Plann Perspect 1991;23:71– 4.
[25] Sayegh MA, Fortenberry JD, Anderson JG, Orr DP. Relationship
quality effects on adolescent chlamydia infections. J Adolesc Health
2005;37:163.e1– e7.
[26] Humphries HO, Bauman KE. Condom use by Norplant users at risk
for sexually transmitted diseases. Sex Transm Dis 1994;21:217–9.
[27] Bankole A, Darroch JE, Singh S. Determinants of trends in condom
use in the United States, 1988 –1995. Fam Plann Perspect 1999;31:
264 –71.

NIH Public Access
Author Manuscript
Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

NIH-PA Author Manuscript

Published in final edited form as:
Perspect Sex Reprod Health. 2014 June ; 46(2): 73–81. doi:10.1363/46e1014.

Dual Method Use Among a Sample of First-Year College Women
Jennifer L. Walsh [assistant professor],
Centers for Behavioral and Preventive Medicine, The Miriam Hospital, Providence; and
Department of Psychiatry and Human Behavior, Alpert Medical School, and Department of
Behavioral and Social Sciences, School of Public Health, Brown University, Providence.
Robyn L. Fielder [postdoctoral fellow],
Center for Integrated Healthcare, Syracuse VA Medical Center, Syracuse, NY.
Kate B. Carey [professor], and
Department of Behavioral and Social Sciences, Program in Public Health, and Center for Alcohol
and Addiction Studies, Brown University.

NIH-PA Author Manuscript

Michael P. Carey [professor]
Centers for Behavioral and Preventive Medicine, The Miriam Hospital, Providence; and
Department of Psychiatry and Human Behavior, Alpert Medical School, and Department of
Behavioral and Social Sciences, School of Public Health, Brown University, Providence.
Jennifer L. Walsh: [email protected]

Abstract
CONTEXT—Dual method use—using one type of contraceptive to reduce the risk of STDs and
another to prevent pregnancy—is effective but understudied. No prior studies have employed an
event-level approach to examining characteristics associated with dual method use among college
women.

NIH-PA Author Manuscript

METHODS—In 12 consecutive monthly surveys conducted in 2009–2010, data on 1,843 vaginal
intercourse events were collected from 296 first-year college women. Women reported on their
use of condoms and hormonal contraceptives during all events. Multilevel regression analysis was
used to assess associations between event-, month- and person-level characteristics and hormonal
use and dual method use.
RESULTS—Women used hormonal contraceptives during 53% of events and condoms during
63%. Dual method use was reported 28% of the time, and only 14% of participants were
consistent users of both methods. The likelihood of dual method use was elevated when sex
partners were friends as opposed to romantic partners or ex-boyfriends, and among women who
had received an STD diagnosis prior to college (odds ratios, 2.5–2.9); it also increased with level
of religiosity (coefficient, 0.8). Dual use was less likely when less reliable methods were used
(odds ratio, 0.2) and when women reported more months of hormonal use (0.8), were older
(coefficient, −4.7) and had had a greater number of partners before college (−0.3).
CONCLUSIONS—A better understanding of the characteristics associated with dual method use
may help in the design of potential intervention efforts.
STDs and unplanned pregnancy affect many young people. In the United States, 20 million
new STD diagnoses are made each year,1 and although teenagers and young adults (15–24-

Walsh et al.

Page 2

NIH-PA Author Manuscript

year-olds) make up only 25% of the sexually active population, they account for 50% of all
gonorrhea infections and 75% of all chlamydia infections.2 In addition, U.S. women aged
18–19 experience a high rate of unplanned pregnancy (162 per 1,000),3 which exceeds rates
in other industrialized nations.4 These consequences of unprotected sexual behavior are
common despite the availability of highly effective contraceptive and preventive methods.
Dual method use involves the use of a contraceptive to reduce pregnancy risk and another
method to reduce the risk of STDs. This combined approach is recommended because
condom use is the most effective method for preventing the spread of STDs,5 whereas
hormonal contraception is the most effective method for pregnancy prevention.6

NIH-PA Author Manuscript

Despite the efficacy of dual method use as a protective strategy, uptake of this practice
remains low; in one review, rates of dual use ranged from 12% among sexually active
women aged 21–25 (2006–2008) to 23% among men and women aged 18–26 and in dating
relationships (2002–2005).7 National data indicate that fewer than one-third of sexually
active, unmarried women aged 15–19 use condoms consistently,8 and rates are even lower
when hormonal contraceptives are used9 and in romantic relationships.10 Increasing dual
method use among 15–19-year-old females is a goal of Healthy People 2020.11 To improve
our understanding about dual method use, for this exploratory study, we examined the
prevalence, use patterns and correlates of dual use among female college students.

BACKGROUND
Gaps in the Literature
The literature on dual method use is limited in several ways. First, most studies have focused
on younger adolescents;12 accordingly, we have limited information about the prevalence of
or characteristics associated with dual method use among adolescents older than 18 and
college students. This is surprising, given that college students are likely to be sexually
active and to engage in serial monogamy,13 and thus have multiple sexual partnerships
during the college years.

NIH-PA Author Manuscript

Second, many studies of the correlates of dual method use have investigated a small number
of variables.14–16 However, sexual behavior is influenced by multiple variables, and is
associated with individual, dyadic, familial, peer and other sociocultural characteristics.17
Research that evaluates a wider range of correlates reflecting a more ecological framework
is needed to better understand sexual and contraceptive behavior.18
Third, most studies have relied on one-time measures involving a long recall period (e.g.,
the past 3–6 months), which undermines the reliability of the data,19 and implicit averaging
across events, which results in less precise assessment. Such methods do not capture
important variability among events, as use of both condoms and hormonal contraceptives
can change over time, and use of either method is often inconsistent. Moreover, features
related to the sexual event itself (e.g., partner type or substance use) may be associated with
dual method use. Event-level studies address these concerns; however, studies limited to
single events are also imperfect, because if that event is not representative, it can distort our
understanding. Hence, event-level studies that employ multiple events are needed, as they

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 3

can provide reliable and precise data that are more representative of a person’s sexual
experiences.20

NIH-PA Author Manuscript

Potential Correlates of Use
A wide range of sociocultural and behavioral characteristics may be associated with
contraceptive use, and these associations have been examined to varying degrees.

NIH-PA Author Manuscript

•Relationship type and duration—Hormonal contraceptive use becomes more frequent,
and condom use less frequent, as relationship duration increases,21–23 and relationship
commitment and duration are negatively associated with dual method use.24,25 Moreover
frequent intercourse (common with steady partners) is negatively related to condom use26
and dual method use.27 In addition, condom use is more common with casual than with
committed partners.28,29 Notably, few studies have considered the likelihood of dual use
with specific types of sexual partners (e.g., acquaintances, friends, ex-boyfriends or new
romantic partners). However, studies have suggested that the types of sexual activities
engaged in and the emotional reactions to sexual encounters differ across types of casual
sexual partners,30,31 and one study found that condom use was more common with exboyfriends, ex-girlfriends and acquaintances than with strangers and friends.30
•Substance use—Drinking, smoking marijuana and cigarette smoking may be related to
dual method use, although results from past research are mixed. Alcohol and marijuana use
are thought to interfere with sexual decision making,32,33 and these behaviors may lead to
decreases in condom use34,35 and dual method use.14 However, one study of college
students found no association between substance use and dual method use.25 Research has
also associated cigarette smoking with higher levels of risky sexual behavior,36 although
smoking in adolescence may be positively associated with consistent condom use in young
adulthood.37
•Hormonal contraceptive use—Research has suggested that as women gain experience
with hormonal methods, they may reduce their condom use (and thus dual method use),
possibly because they are more confident that their hormonal contraceptive will protect
against pregnancy.26

NIH-PA Author Manuscript

•Sexual history and risk perceptions—One study suggested that women who have
partners they perceive as risky may be more likely than others to use dual methods,25 while
another found that dual use was negatively correlated with perceptions of STD risk.16 Other
research has found that women who have had an STD are more likely than others to use
condoms,38 possibly because they place greater importance on protection. Some studies
have found that young people who have had a greater number of sexual partners are more
likely than those who have had fewer partners to use dual methods,39,40 but research has
also suggested that condom use declines more rapidly within relationships for those who
have had more past partners.23 An additional element that may relate to risk perception is
use of less reliable contraceptive methods. The diaphragm, sponge, fertility awareness and
withdrawal are less effective than condoms and hormonal contraceptives for both pregnancy
and STD prevention;6,41 however, women using these methods may believe they are

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 4

NIH-PA Author Manuscript

adequately protected.42 Most prior studies of dual method use have either not considered
these methods or included them in the definition of dual method use despite their lower
effectiveness.12,14,16
•Parent and peer characteristics—One study found that positive parental attitudes
toward condom use and birth control were associated with increased levels of dual method
use.16 Similarly, discussing safer sex with parents has been correlated with dual use;12,43
however, studies have not investigated the role of communication with peers. Parental
connectedness appears to be protective against sexual risk behavior and pregnancy,44 so we
might expect it to be positively associated with dual method use, although no studies have
specifically examined this.
•Personality—Impulsivity and sensation-seeking are both related to risky sexual
behavior,45 and one study found that impulsivity was negatively associated with dual use.46
In contrast, conscientiousness has been found to be positively associated with condom use,47
but its association with dual method use has not been explored.

NIH-PA Author Manuscript

•Demographic characteristics—Past research has suggested that the prevalence of dual
use is higher among adolescents than among adults,7,39 and higher among blacks than
among whites or Latinas.15 Religiosity is commonly regarded as protective against sexual
activity in general,48 but studies have found mixed results regarding its association with
condom use,48,49 and the possible correlation between religiosity and dual method use has
not been examined. A final demographic correlate is socioeconomic status; one study
showed that teenagers from higher status groups were more likely than others to use
condoms,50 and this variable has been assessed in some dual method studies.16
The Current Study

NIH-PA Author Manuscript

This exploratory study addresses two questions: What is the prevalence of dual method use
in a sample of first-year college women? And what characteristics are associated with dual
use in this sample? We used event-level data from multiple sexual events, and considered a
wide variety of event-, month- and person-level characteristics that may be associated with
dual method use, many of which have rarely or never been considered previously.
Consistent with a behavioral ecological framework,18 we included proximal variables
related to the sex partner and characteristics of the sexual event—often neglected in previous
studies—as well as more distal ones such as family and peer characteristics. We also
considered two categories of risk behavior—substance use and past sexual experience—
which are important elements in an ecological framework.18

METHODS
Participants and Procedures
Women came from a pool of 483 first-year students who were attending a private university
in upstate New York and who were participating in a yearlong study of health behaviors and
relationships. The larger study, conducted between September 2009 and September 2010,
explored a variety of health behaviors (e.g., substance use, diet, exercise, sleep), as well as

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 5

NIH-PA Author Manuscript

sexual behavior and psychosocial adjustment.51–53 The full sample constituted 26% of firstyear female students for the fall 2009 semester. The 296 women included in the current
study reported at least one episode of vaginal intercourse with a romantic or casual partner
during their first year of college.
The university’s institutional review board approved all study procedures. Participants were
recruited via a mass mailing sent to first-year female students. Campus flyers, word of
mouth and the psychology department participant pool bolstered recruitment. Most
participants (61%) heard about the study through the mass mailing, 28% signed up through
the department pool, and 11% responded to flyers or word of mouth. Interested students
attended an orientation session, after which they provided informed consent and completed
the initial survey. Subsequently, participants completed monthly online assessments for one
year; surveys were completed during the first week of each month, and reports covered the
previous month. Participants received $10–20 for each survey, depending on its length.
Measures

NIH-PA Author Manuscript

• Event-level variables—At each monthly survey, women who said they had had oral or
vaginal sex during the past month reported on their most recent encounter involving oral,
vaginal or anal sex with both a romantic partner and a casual partner. (A romantic partner
was defined as “someone whom you were dating or in a romantic relationship with at the
time of the physical intimacy”; a casual partner was “someone whom you were not dating or
in a romantic relationship with at the time of the physical intimacy, and there was no mutual
expectation of a romantic commitment; some people call these hookups.”) Thus, each
participant could describe 0–2 encounters per month, or 0–24 total. Only reports of vaginal
sex in the preceding month were included in this analysis.

NIH-PA Author Manuscript

For vaginal intercourse events, participants reported all contraceptive methods they had
used. Response options were nothing; male condom; pill, patch or ring; withdrawal (“pulling
out”); injectable; female condom; IUD; diaphragm, cervical cap or sponge; fertility
awareness (calendar, mucus, basal body temperature); and other. Women were coded as
using a condom if they reported male or female condom use. They were coded as using
another reliable contraceptive if they had used the pill, patch, ring, injectable or IUD; no
women in our sample reported IUD use, so this category is henceforth referred to as
“hormonal contraceptive use.” Women were coded as using a less reliable method if they
had used withdrawal; a diaphragm, cap or sponge; or fertility awareness.6
For events with romantic partners, relationship duration was dichotomized into one month or
less or longer than a month. The one-month cutoff was chosen because condom use begins
to decline within weeks of beginning a new relationship.10 For events with casual partners,
participants were asked to identify their partner; response options were a stranger, an
acquaintance, a friend, an ex-boyfriend and other. Answers of “other” were rare (2%) and
were coded as missing.
Participants also reported whether they had drunk alcohol or used marijuana before each
event.

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 6

NIH-PA Author Manuscript

• Month-level variables—Some variables were assessed on a monthly basis but were not
linked directly to the sexual events reported. Because measures were collected over one
year, we included the month of data collection in models, ranging from 2 (October 2009) to
13 (September 2010). Participants who were involved in romantic relationships reported
their relationship duration in months. Women indicated the number of days in the past
month during which they had engaged in binge drinking (consuming four or more drinks on
one occasion) or had smoked marijuana. They were also asked whether they had smoked a
cigarette during the past month; those who had smoked reported the average number of
cigarettes per day. Because of low rates of cigarette smoking, a dummy variable was created
indicating whether participants had smoked at least one cigarette a day, on average.
Participants reported the number of times they had had vaginal intercourse with romantic
and casual partners during the past month. We created separate counts of monthly
intercourse events by type of partner (excluding the event under analysis). Similarly,
participants reported the number of romantic and casual partners they had had intercourse
with in the past month; these counts also excluded the event under analysis. Finally, we
created a variable indicating women’s average number of months of hormonal contraceptive
use; this included only months after women had enrolled in college.

NIH-PA Author Manuscript

• Person-level variables—These variables were assessed only once during the year.
Several items concerned sexual history. At baseline, participants indicated the number of
partners they had had vaginal intercourse with, whether they had ever received an STD
diagnosis and whether they had ever been pregnant prior to enrolling. In April 2010,
participants answered one item assessing perceived STD risk: “What do you think your
chances are of getting an STD, such as gonorrhea or genital herpes?” Responses were rated
on a scale from 1 (no chance) to 5 (very high).

NIH-PA Author Manuscript

Scales were used to assess parental and peer characteristics. In October 2009, participants
completed eight items, adapted from a subscale of the Parenting Style Index,54 indicating
parental connectedness (e.g., “I can count on my parents to help me out if I have some kind
of problem”). Responses were scored on a scale from 1 (strongly disagree) to 4 (strongly
agree) and were averaged to create a total score (Cronbach’s alpha, 0.91). In August 2010,
participants answered two items about how their parents would feel about their using birth
control at this point in their life; responses were rated on a scale from 1 (strongly
disapprove) to 5 (strongly approve).55 Items were averaged for participants reporting on
both parents (Cronbach’s alpha, 0.75); single items were used for those reporting on only
one parent. In March 2010, participants completed items from the Parent-Adolescent
Communication Scale56 indicating how often since starting college they and their parents
had discussed five sexual topics (e.g., protecting themselves from pregnancy and how to use
condoms); the scale ranged from 0 (never) to 4 (often). These items were averaged to create
a total score (Cronbach’s alpha, 0.91). Finally, in March 2010, participants indicated how
often in the past month they had discussed three sexual topics with peers: having sex,
protection against STDs and protection against pregnancy; responses were scored on a scale
from 1 (never) to 6 (nearly every day). Items were adapted from a peer alcohol
communication assessment57 and were averaged to create a total score (Cronbach’s alpha,
0.82).
Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 7

NIH-PA Author Manuscript

Three scales assessed personality. At baseline, impulsivity and sensation-seeking were
measured using six items each from subscales of the Impulsiveness-Monotony Avoidance
Scale.58 Participants indicated how well each item (e.g., “I often throw myself too hastily
into things” and “I like doing things just for the thrill of it”) applied to them, using a Likert
scale from 1 (not at all like me) to 4 (very much like me). Scores were averaged to create a
total score (Cronbach’s alpha, 0.82 for each scale). Finally, in June 2010, participants
responded to items from the Ten-Item Personality Inventory59 indicating how strongly they
agreed that various traits represented them; responses were scored on a scale from 1
(strongly disagree) to 7 (strongly agree). Two items represented conscientiousness
(“dependable” and “self-disciplined”) and were averaged to create a total score (Cronbach’s
alpha, 0.42).

NIH-PA Author Manuscript

Several demographic variables were also assessed. A dummy variable was created to
indicate whether participants were older than 18 at baseline, and two other variables
indicated whether participants identified themselves as white, black or Asian, or as Latina.
Socioeconomic status was assessed using a 10-point ladder,60 on which participants ranked
their family relative to other U.S. families. Finally, participants reported the extent to which
they considered themselves religious (from “not religious” to “very religious”) and their
frequency of attending religious services (from “never” to “more than once a week”). These
items were averaged, and higher scores on a 0–3 scale indicated greater religiosity
(Cronbach’s alpha, 0.80).
Analysis
Completion rates for monthly surveys ranged from 82% (in month 11) to 100% (in month
1); on average, participants completed 11.8 months of data collection (standard deviation,
2.2). To maintain the entire sample, we used multiple imputation to replace missing
values.61 Multiple imputation is a method for dealing with missing data that avoids biases
associated with the use of only complete cases or with single imputations.62 We imputed
100 complete data sets using Mplus 7,63 and all study variables were included in the
imputation. Analyses were conducted with all 100 data sets, and parameter estimates were
pooled using the imputation algorithms in Mplus.

NIH-PA Author Manuscript

We used multilevel modeling in Mplus 7 to analyze the data. A total of 1,843 sexual events
were reported by 296 participants. Given that hormonal contraceptive use is unlikely to vary
between events occurring in the same month, we first explored associations between
hormonal use and month-level and person-level characteristics. Next, we examined
associations between event-, month- and person-level characteristics and condom use in 977
events reported by 181 women in which hormonal contraceptives were also used.
Coefficients for variables that were highly nonsignificant (Z<1.00) were constrained to zero
to increase model parsimony and stabilize estimates.64 Odds ratios (from logistic regression
analyses at the event and month levels), unstandardized betas (from linear regression
analyses at the person level) and 95% confidence intervals are reported throughout.

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 8

RESULTS
NIH-PA Author Manuscript

Sample Characteristics
The majority of vaginal intercourse events occurred with established romantic partners
(53%); events with new romantic partners (22%) and friends (15%) were the next most
common (Table 1). Participants reported using alcohol prior to 20% of events, and
marijuana before 7%. For 30% of intercourse events, women had used a less reliable
contraceptive method.
The average duration of romantic relationships was nine months. Binge drinking and
marijuana use occurred 2–3 times per month, on average, but cigarette smoking was
relatively uncommon. Women reported a monthly average of 5.5 intercourse events with
romantic partners and 0.4 events with casual partners; most women did not report other
romantic or casual partners. On average, women who used hormonal contraceptives did so
for four months during college.

NIH-PA Author Manuscript

The great majority of participants (96%) were age 18; the mean age was 18.1 (standard
deviation, 0.2—not shown). Most women were white (71%); the remainder were black
(13%), Asian (8%) or of other race (7%). Overall, 11% were Latina. On average,
participants reported being middle- to upper-middle class, and had relatively low levels of
religiosity.
Participants perceived themselves to be at relatively low risk of STD infection, despite
reporting two sexual partners, on average, prior to college; few women reported a history of
either STD diagnosis or pregnancy. On average, women reported a high level of
connectedness with their parents, perceived their parents to have neutral or positive attitudes
toward birth control, and reported communicating about sex with both their parents and
peers relatively infrequently. Finally, women reported moderate levels of impulsivity and
sensation-seeking, and a high level of conscientiousness.
Patterns of Contraceptive Use

NIH-PA Author Manuscript

Across 1,843 sexual intercourse events, women reported using a variety of contraceptive
methods (Table 2). Male condoms were used in 63% of events; the pill, patch or ring in
53%; withdrawal in 30%; and no method in 6%. Other methods were used rarely.
Multiple contraceptive methods were used in 45% of events. Condoms and hormonal
methods (the combination of primary interest in this study) were used together in 28% of
events, condoms and less reliable methods in 13%, and hormonal and less reliable methods
in 14%. In 5% of all cases, a condom, a hormonal method and a less reliable method were
used in combination.
Of the 296 women who engaged in intercourse over the course of the study, 39% did not
report any hormonal contraceptive use, while 34% reported such use during all intercourse
events. The remaining 27% reported inconsistent use of hormonal methods across the year;
these women either initiated use (13%), stopped use (7%) or reported other patterns of use
(6%).

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 9

NIH-PA Author Manuscript

Forty-six percent of women were consistent condom users across all reported events, while
11% never used condoms; the remaining 43% reported inconsistent condom use. Only 14%
of women were consistent dual method users; 53% never used dual methods, and the
remaining 33% used dual methods inconsistently.
Correlates of Method Use
•Hormonal methods—Our multilevel model (Table 3) showed that women were more
likely to use a hormonal contraceptive if they reported more frequent intercourse with
romantic partners (odds ratio, 1.1) and perceived that their parents had more positive
attitudes toward birth control (coefficient, 1.2). In contrast, women had a reduced likelihood
of reporting hormonal use if they were older than 18 (−4.7), black (−4.3) or more religious
(−1.0). This model explained 18% of the event-level variance and 25% of the person-level
variance.

NIH-PA Author Manuscript

•Dual methods—Women were more likely to report dual method use when their partner
was a friend rather than an established romantic partner (odds ratio, 2.5). In analyses that
compared all partner categories (not shown), women were more likely to be dual users when
their partner was a friend rather than a new romantic partner (2.5; 95% confidence interval,
1.0–6.1) or an ex-boyfriend (2.8; 95% confidence interval, 1.2–6.6); differences were not
found between any other categories.
Women had a reduced likelihood of reporting dual method use if they had used a less
reliable contraceptive method (odds ratio, 0.2) or had more months of experience with
hormonal methods (0.8). At the person level, women were less likely to report dual use if
they were older than 18 (coefficient, −4.7) or had had a greater number of past partners
(−0.3). They were more likely to have been dual users if they reported greater religiosity
(0.8) or had received an STD diagnosis prior to entering college (2.9). This model explained
29% of the event-level variance and 20% of the person-level variance.

DISCUSSION

NIH-PA Author Manuscript

Hormonal contraceptives and condoms were used together in only a quarter of intercourse
events reported by participants; half of the women said they had never used dual methods,
and a third reported inconsistent dual use. The inconsistency of dual method use among our
participants indicates the need to examine associated characteristics using event-level, timevarying data rather than cross-sectional data. The prevalence of dual method use in our
sample appears to be similar to that in other populations.7 Given the high rates of STDs and
unplanned pregnancy among women aged 18–19,2,3 and the ubiquity of serial monogamy
during the college years,65 increasing the rates of dual method use among college women
should be a goal for health educators and providers.
Our study showed that dual use declined as experience with hormonal contraceptives
increased—by 20% with each additional month of use. One previous study26 addressed the
decline in condom use with increasing experience with hormonal contraceptives; however,
these researchers noted a decline primarily among injectable users but not among pill users.
We found a negative association between hormonal experience and dual method use, even

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 10

NIH-PA Author Manuscript

though the great majority of hormonal users in our study were using the pill, patch or ring.
Women who used a less reliable contraceptive in conjunction with a hormonal method were
also relatively unlikely to use a condom. Women may feel more confident about pregnancy
prevention without a condom as their experience with hormonal methods increases, or if
they use withdrawal or natural family planning methods in addition to a hormonal
contraceptive. However, these less effective methods do not protect against STD
transmission.

NIH-PA Author Manuscript

Women’s number of past sexual partners was inversely related to their reports of dual
method use. This finding contrasts with results from other studies, in which young people
with more recent partners were more likely to use condoms.39,40 The reduced likelihood of
dual use women with more previous partners may be due to the tendency for condom use to
decline more rapidly in each successive sexual relationship.23 Although relatively few
women in our sample had received an STD diagnosis, those who had had an STD were more
likely than others to engage in dual method use, consistent with theories that invoke
perceived risk as a determinant of health behaviors.66 Interestingly, however, our measure of
STD risk perception was not associated with dual use. Women with past diagnoses may
perceive themselves as being at lower risk as a result of their condom use. Future research
should employ multiitem measures of risk perceptions.
Religiosity had a complex relationship with contraceptive use in our sample. Although
women who were more religious were less likely than others to use hormonal
contraceptives, among those who used these methods, religiosity was positively associated
with dual method use. The role of religiosity should be further explored in future studies.
Compared with women who were having sex with romantic partners (both short- and longterm), those having sex with friends were more likely to report dual method use. Previous
studies have found that dual use is less common with romantic than with casual
partners,24,25 but they did not examine associations between specific partner types and dual
method use. Although “hookups” (casual sexual encounters, which often occur with
friends30) are often framed in a negative light, this finding suggests that some hookup
experiences may involve protection against both pregnancy and STDs.

NIH-PA Author Manuscript

A number of variables investigated in our study, including sensation-seeking,
conscientiousness, and parental and peer communication about sex, were not associated with
dual method use. Surprisingly, substance use—at either the month or the event level—was
not associated with dual use, in contrast to another study’s finding that situational alcohol
use was negatively related to dual method use among men and women aged 20–23.14 Future
studies might examine gender-specific associations between situational substance use and
dual method use, ideally using event-level data.
Limitations and Future Directions
Several limitations of the current study suggest directions for future research. First, our data
came from first-year female students at one university; generalizability to other settings and
populations is not certain. Future studies should collect data from males and from a range of
other institutions. In addition, although use of event-level measures is a major strength, we

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 11

NIH-PA Author Manuscript

cannot determine causal relationships from our data; some variables could be either
predictors or outcomes of dual method use. Moreover, a number of our variables were
assessed only once; future studies using event-level data should assess some of these
variables (such as religiosity, parent and peer characteristics, and risk perceptions) over
time, as more precise measurement may strengthen the assessment of their relationship with
dual method use. Finally, our study did not assess some proximal variables, such as attitudes
and intentions related to dual method use, that have been studied more commonly than
event-level features in the past.16,40 Given the inconsistency of dual method use, future
research should consider both event-level characteristics and constructs from ecological and
health behavior theories.67 Ideally, this research should be based on specific theoretical
models; our exploratory study suggests some variables that these models should incorporate.
Conclusions

NIH-PA Author Manuscript

This study identified a number of characteristics that are associated with dual method use,
which occurred in only a quarter of reported intercourse events. A better understanding of
such characteristics may aid in intervention design; if our results are borne out by more
generalizable studies, the implication may be that women need to be counseled on the
importance of maintaining use of condoms as their sexual relationships become more
serious, as they gain experience with hormonal contraceptives and even when they are also
using less reliable methods, such as withdrawal. Our findings demonstrate the importance of
considering event-level and time-varying characteristics that may be associated with dual
method use, in addition to the person-level characteristics that are more commonly
considered. Future research should collect further data related to event-level characteristics
to begin to inform potential intervention efforts. Moreover, longitudinal research is needed
to identify predictors of dual method use.

Acknowledgments
This research was supported by grant R21-AA018257 from the National Institute on Alcohol Abuse and
Alcoholism.

REFERENCES
NIH-PA Author Manuscript

1. Centers for Disease Control and Prevention (CDC). STD Trends in the United States: 2011 National
Data for Chlamydia, Gonorrhea and Syphilis. Atlanta: CDC; 2013.
2. CDC. Sexually Transmitted Disease Surveillance 2009. Atlanta: CDC; 2010.
3. Finer LB. Unintended pregnancy among U.S. adolescents: accounting for sexual activity. Journal of
Adolescent Health. 2010; 47(3):312–314. [PubMed: 20708573]
4. Singh S, Darroch JE. Adolescent pregnancy and childbearing: levels and trends in developed
countries. Family Planning Perspectives. 2000; 32(1):14–23. [PubMed: 10710702]
5. Holmes KK, Levine R, Weaver M. Effectiveness of condoms in preventing sexually transmitted
infections. Bulletin of the World Health Organization. 2004; 82(6):454–461. [PubMed: 15356939]
6. RA, Hatcher; J, Trussell; AL, Nelson, editors. Contraceptive Technology. 19th rev. ed. New York:
Ardent Media; 2007.
7. Williams RL, Fortenberry JD. Dual use of long-acting reversible contraceptives and condoms
among adolescents. Journal of Adolescent Health. 2013; 52(4, Suppl.):S29–S34. [PubMed:
23535054]
8. Chandra A, et al. Fertility, family planning, and reproductive health of U.S. women: data from the
2002 National Survey of Family Growth. Vital and Health Statistics. 2005 Series 23, No. 25.
Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 12

NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript

9. Cooper ML, Agocha VB, Powers AM. Motivations for condom use: Do pregnancy prevention goals
undermine disease prevention among heterosexual young adults? Health Psychology. 1999; 18(5):
464–474. [PubMed: 10519462]
10. Fortenberry JD, et al. Condom use as a function of time in new and established adolescent sexual
relationships. American Journal of Public Health. 2002; 92(2):211–213. [PubMed: 11818293]
11. U.S. Department of Health and Human Services (DHHS). Healthy People 2020: The Road Ahead.
Washington, DC: DHHS; 2010.
12. Crosby RA, et al. Correlates of using dual methods for sexually transmitted diseases and pregnancy
prevention among high-risk African-American female teens. Journal of Adolescent Health. 2001;
28(5):410–414. [PubMed: 11336871]
13. Corbin WR, Fromme K. Alcohol use and serial monogamy as risks for sexually transmitted
diseases in young adults. Health Psychology. 2002; 21(3):229–236. [PubMed: 12027028]
14. Bailey JA, et al. Romantic relationship characteristics and alcohol use: longitudinal associations
with dual method contraception use. Journal of Adolescent Health. 2012; 50(5):450–455.
[PubMed: 22525107]
15. Eisenberg DL, et al. Correlates of dual-method contraceptive use: an analysis of the National
Survey of Family Growth (2006–2008). Infectious Diseases in Obstetrics and Gynecology. 2012
16. Sieving RE, et al. Adolescent dual method use: relevant attitudes normative beliefs and selfefficacy. Journal of Adolescent Health. 2007; 40(3):275. e15–275.e22, , [PubMed: 17321430]
17. Kotchick BA, et al. Adolescent sexual risk behavior: a multi-system perspective. Clinical
Psychology Review. 2001; 21(4):493–519. [PubMed: 11413865]
18. Hovell MF, et al. A behavioral-ecological model of adolescent sexual development: a template for
AIDS prevention. Journal of Sex Research. 1994; 31(4):267–281.
19. Schroder KE, Carey MP, Vanable PA. Methodological challenges in research on sexual risk
behavior. II. accuracy of self-reports. Annals of Behavioral Medicine. 2003; 26(2):104–123.
[PubMed: 14534028]
20. Weinhardt LS, Carey MP. Does alcohol lead to sexual risk behavior? Findings from event-level
research. Annual Review of Sex Research. 2000; 11(1):125–157.
21. Civic D. College students’ reasons for nonuse of condoms within dating relationships. Journal of
Sex & Marital Therapy. 2000; 26(1):95–105. [PubMed: 10693119]
22. Hammer JC, et al. When two heads aren’t better than one: AIDS risk behavior in college-age
couples. Journal of Applied Social Psychology. 1996; 26(5):375–397.
23. Ku L, Sonenstein FL, Pleck JH. The dynamics of young men’s condom use during and across
relationships. Family Planning Perspectives. 1994; 26(6):246–251. [PubMed: 7867771]
24. Wilson EK, Koo HP. Associations between low-income women’s relationship characteristics and
their contraceptive use. Perspectives on Sexual and Reproductive Health. 2008; 40(3):171–179.
[PubMed: 18803799]
25. Poppen PJ, Reisen CA. Women’s use of dual methods of sexual self-protection. Women & Health.
2000; 30(2):53–66. [PubMed: 10881758]
26. Sayegh MA, et al. The developmental association of relationship quality hormonal contraceptive
choice and condom non-use among adolescent women. Journal of Adolescent Health. 2006; 39(3):
388–395. [PubMed: 16919801]
27. Weisman CS, et al. Consistency of condom use for disease prevention among adolescent users of
oral contraceptives. Family Planning Perspectives. 1991; 23(2):71–74. [PubMed: 2060614]
28. Misovich SJ, Fisher JD, Fisher WA. Close relationships and elevated HIV risk behavior: evidence
and possible underlying psychological processes. Review of General Psychology. 1997; 1(1):72–
107.
29. Noar, SM., et al. Safer sex and sexually transmitted infections from a relationship perspective, in.
In: Harvey, JH.; Wenzel, A.; Sprecher, S., editors. Handbook of Sexuality in Close Relationships.
Mahwah, NJ: Lawrence Erlbaum; 2004. p. 519-544.
30. Lewis MA, et al. Predictors of hooking up sexual behaviors and emotional reactions among U.S.
college students. Archives of Sexual Behavior. 2012; 41(5):1219–1229. [PubMed: 21796484]

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 13

NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript

31. Furman W, Shaffer L. Romantic partners, friends, friends with benefits and casual acquaintances
as sexual partners. Journal of Sex Research. 2011; 48(6):554–564. [PubMed: 21128155]
32. Steele CM, Josephs RA. Alcohol myopia: its prized and dangerous effects. American Psychologist.
1990; 45(8):921–933. [PubMed: 2221564]
33. Lane SD, et al. Acute marijuana effects on human risk taking. Neuropsychopharmacology. 2005;
30(4):800–809. [PubMed: 15775958]
34. Scott-Sheldon LAJ, Carey MP, Carey KB. Alcohol and risky sexual behavior among heavy
drinking college students. AIDS and Behavior. 2010; 14(4):845–853. [PubMed: 18648928]
35. Bryan AD, Schmiege SJ, Magnan RE. Marijuana use risky sexual behavior among high-risk
adolescents: trajectories, risk factors and event-level relationships. Developmental Psychology.
2012; 48(5):1429–1442. [PubMed: 22390666]
36. Duncan SC, Strycker LA, Duncan TE. Exploring associations in developmental trends of
adolescent substance use and risky sexual behavior in a high-risk population. Journal of
Behavioral Medicine. 1999; 22(1):21–34. [PubMed: 10196727]
37. Guo J, et al. Developmental relationships between adolescent substance use and risky sexual
behavior in young adulthood. Journal of Adolescent Health. 2002; 31(4):354–362. [PubMed:
12359381]
38. Soskolne V, et al. Condom use with regular and casual partners among women attending family
planning clinics. Family Planning Perspectives. 1991; 23(5):222–225. [PubMed: 1743275]
39. Harvey SM, Henderson JT, Branch MR. Protecting against both pregnancy and disease: predictors
of dual method use among a sample of women. Women & Health. 2004; 39(1):25–43. [PubMed:
15002881]
40. Pack RP, et al. Psychosocial correlates of dual methods for contraception and STI protection in
urban adolescents. ISRN Obstetrics and Gynecology. 2011
41. CDC. Sexually transmitted diseases treatment guidelines, 2010. Morbidity and Mortality Weekly
Report. 2012; 59(RR-12)
42. O’Sullivan LF, et al. A cognitive analysis of college students’ explanations for engaging in
unprotected sexual intercourse. Archives of Sexual Behavior. 2010; 39(5):1121–1131. [PubMed:
19365717]
43. Santelli JS, et al. The use of condoms with other contraceptive methods among young men and
women. Family Planning Perspectives. 1997; 29(6):261–267. [PubMed: 9429871]
44. Miller BC. Family influences on adolescent sexual and contraceptive behavior. Journal of Sex
Research. 2002; 39(1):22–26. [PubMed: 12476252]
45. Charnigo R, et al. Sensation seeking and impulsivity: combined associations with risky sexual
behavior in a large sample of young adults. Journal of Sex Research. 2013; 50(5):480–488.
[PubMed: 22456443]
46. Sales JM, et al. Differences between dual-method and non?dual-method protection use in a sample
of young African American women residing in the southeastern United States. Archives of
Pediatrics & Adolescent Medicine. 2010; 164(12):1125–1131. [PubMed: 21135341]
47. Hoyle RH, Fejfar MC, Miller JD. Personality and sexual risk taking: a quantitative review. Journal
of Personality. 2000; 68(6):1203–1231. [PubMed: 11130738]
48. Zaleski EH, Schiaffino KM. Religiosity and sexual risk-taking behavior during the transition to
college. Journal of Adolescence. 2000; 23(2):223–227. [PubMed: 10831144]
49. McCree DH, et al. Religiosity and risky sexual behavior in African-American adolescent females.
Journal of Adolescent Health. 2003; 33(1):2–8. [PubMed: 12834991]
50. Abma, JC. Sexual activity among teenagers in the United States, in. In: Weissberg, RP., et al.,
editors. Long-Term Trends in the Well-Being of Children and Youth: Issues in Children’s and
Families’ Lives. Washington, DC: Child Welfare League of America; 2003. p. 103-128.
51. Fielder RL, Carey KB, Carey MP. Are hookups replacing romantic relationships? A longitudinal
study of first-year female college students. Journal of Adolescent Health. 2013; 52(5):657–659.
[PubMed: 23298995]
52. Walsh JL, et al. Changes in women’s condom use over the first year of college. Journal of Sex
Research. 2013; 50(2):128–138. [PubMed: 22235757]

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 14

NIH-PA Author Manuscript
NIH-PA Author Manuscript

53. Fielder RL, Predictors of sexual, hookups: a theory-based, et al. prospective study of first-year
college women. Archives of Sexual Behavior. 2013; 42(8):1425–1441. [PubMed: 23657811]
54. Steinberg L, et al. Impact of parenting practices on adolescent achievement: authoritative parenting
school involvement and encouragement to succeed. Child Development. 1992; 63(5):1266–1281.
[PubMed: 1446552]
55. Harris, KM., et al. [accessed June 25, 2013] The National Longitudinal Study of Adolescent
Health: research design. 2003. ,
56. Sales JM, et al. Validation of a parent-adolescent communication scale for use in STD/HIV
prevention interventions. Health Education & Behavior. 2008; 35(3):332–345. [PubMed:
17200099]
57. Real K, Rimal RN. Friends talk to friends about drinking: exploring the role of peer
communication in the theory of normative social behavior. Health Communication. 2007; 22(2):
169–180. [PubMed: 17668996]
58. Schalling, D. Psychopathy-related personality variables and the psychophysiology of socialization,
in. In: Hare, RD.; Schalling, D., editors. Psychopathic Behavior: Approaches to Research. New
York: Wiley; 1978. p. 85-105.
59. Gosling SD, Rentfrow PJ, Swann WB Jr. A very brief measure of the Big-Five personality
domains. Journal of Research in Personality. 2003; 37(6):504–528.
60. Adler NE, et al. Relationship of subjective and objective social status with psychological and
physiological functioning: preliminary data in healthy white women. Health Psychology. 2000;
19(6):586–592. [PubMed: 11129362]
61. Schafer, JL. Analysis of Incomplete Multivariate Data. New York: Chapman and Hall; 1997.
62. Schafer JL. Multiple imputation: a primer. Statistical Methods in Medical Research. 1999; 8(1):3–
15. [PubMed: 10347857]
63. Muthén, LK.; Muthén, BO. Mplus User’s Guide. seventh ed. Los Angeles: Muthén & Muthén;
2012.
64. Bentler PM, Mooijaart A. Choice of structural model via parsimony: a rationale based on
precision. Psychological Bulletin. 1989; 106(2):315–317. [PubMed: 2678203]
65. MacDonald NE, et al. High-risk STD/HIV behavior among college students. Journal of the
American Medical Association. 1990; 263(23):3155–3159. [PubMed: 2348524]
66. Rosenstock IM. The health belief model and preventive health behavior. Health Education &
Behavior. 1974; 2(4):354–386.
67. Fisher, JD.; Fisher, WA. Theoretical approaches to individual-level change in HIV risk behavior,
in. In: Peterson, JL.; DiClemente, RJ., editors. Handbook of HIV Prevention, AIDS Prevention,
and Mental Health. New York: Kluwer Academic/Plenum; 2000. p. 3-55.

NIH-PA Author Manuscript
Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 15

TABLE 1

NIH-PA Author Manuscript

Selected event-, month- and person-level characteristics of vaginal intercourse events among first-year college
women attending a private university, New York State, 2009–2010
Characteristic

% or mean
(N=1,843)

Event level
Partner type
Stranger

Acquaintance

Friend

Ex-boyfriend

New romantic (≤1 month)

Established romantic (>1 month)

Used alcohol before intercourse

Used marijuana before intercourse

Used less reliable contraceptive†

30
(N=1,843)

Month level

NIH-PA Author Manuscript

Month of data collection (range, 2–13)

7.5 (3.5)

Months of romantic relationship (range, 0–53)

9.1 (12.1)

No. of days engaged in binge drinking (range, 0–14)

2.8 (3.4)

No. of days used marijuana (range, 0–21)

2.4 (5.1)

Smoked cigarettes

No. of intercourse events‡
With romantic partner (range, 0–31)

5.5 (7.0)

With casual partner (range, 0–7)

0.4 (1.2)

No. of intercourse partners (range, 0–2)‡
Romantic

0.1 (0.3)

Casual

0.1 (0.3)

Months of hormonal contraceptive use (range, 1–12)§

4.2 (2.8)

Person level

(N=296)

Aged >18

Race

NIH-PA Author Manuscript

White

Black

Asian

Other

Latina ethnicity

Family socioeconomic status (range, 1–10)

6.3 (1.6)

Religiosity (range, 0–3)

0.9 (0.7)

No. of intercourse partners before college (range, 0–9)

2.4 (2.4)

Perceived STD risk (range, 1–5)

2.0 (0.9)

Ever had STD diagnosis before college

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 16

Characteristic

% or mean

Ever pregnant before college

NIH-PA Author Manuscript

Parental connectedness (range, 1–4)

2
3.4 (0.5)

Parental attitude toward birth control (range, 1–5)

3.5 (1.1)

Parental communication about sex (range, 0–4)

1.7 (0.7)

Peer communication about sex (range, 1–6)

2.5 (1.1)

Impulsivity (range, 1–4)

2.2 (0.6)

Sensation-seeking (range, 1–4)

2.9 (0.6)

Conscientiousness (range, 1–7)

5.2 (1.1)

†

Withdrawal, diaphragm, cervical cap, sponge or fertility awareness.

‡

Excludes the current event or partner.

Includes only women who reported use (181 women across 977 months).

Notes: Data for which no ranges are shown are percentages. Figures in parentheses are standard deviations.

NIH-PA Author Manuscript
NIH-PA Author Manuscript
Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 17

TABLE 2

NIH-PA Author Manuscript

Percentage of vaginal intercourse events reported by first-year college women, by contraceptive method used
Method

Condom

Male

Female

0.1

Hormonal method

Pill/patch/ring

Injectable

0.3

IUD

Less reliable method

Withdrawal

Fertility awareness

Diaphragm

0.3

No method

NIH-PA Author Manuscript

Multiple methods

Condom plus hormonal method

Condom plus less reliable method

Hormonal plus less reliable method

≥2 methods

≥3 methods

NIH-PA Author Manuscript
Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 18

TABLE 3

NIH-PA Author Manuscript

Odds ratios from logistic regression analyses assessing associations between event-level and month-level
characteristics of first-year college women’s vaginal intercourse events and hormonal contraceptive use or
dual method use, and unstandardized coefficients from linear regression analyses assessing associations
between person-level characteristics and such use
Characteristic

Hormonal use

Dual use

Stranger

‡

Event level
Partner type

Acquaintance

‡

Friend

2.47 (1.001–6.11)*

Ex-boyfriend

0.42 (0.09–1.88)

New romantic

‡

Established romantic (ref)

1.00

NIH-PA Author Manuscript

Used alcohol before intercourse

1.68 (0.81–3.50)

Used marijuana before intercourse

‡

Used less reliable contraceptive

0.24 (0.11–0.52)***

Month of data collection

1.10 (0.99–1.21)†

1.11 (0.97–1.28)

Months of romantic relationship

‡

No. of days engaged in binge drinking

‡

1.05 (0.95–1.15)

No. of days used marijuana

1.06 (0.96–1.17)

‡

Smoked cigarettes

0.58 (0.25–1.36)

‡

With romantic partner

1.11 (1.05–1.17)***

0.97 (0.93–1.01)

With casual partner

‡

Romantic

‡

2.65 (0.96–7.32)†

Casual

‡

0.80 (0.66–0.96)*

−4.71 (−8.72 to −0.69)*

−4.66 (−8.76 to −0.55)*

White (ref)

1.00

Black

−4.25 (−6.60 to −1.90)***

‡

Asian

‡

1.59 (−0.21 to 3.40)†

Month level

No. of intercourse events

No. of intercourse partners

Months of hormonal contraceptive use

NIH-PA Author Manuscript

Person level
Age >18
Race

Latina ethnicity

‡

Family socioeconomic status

0.33 (−0.11 to 0.77)

‡

Religiosity

−0.96 (−1.88 to −0.04)*

0.75 (0.01–1.49)*

No. of intercourse partners before college

‡

−0.34 (–0.59 to –0.08)**

Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

Walsh et al.

Page 19

NIH-PA Author Manuscript

Characteristic

Hormonal use

Dual use

Perceived STD risk

−0.47 (−1.34 to 0.40)

‡

Ever had STD diagnosis before college

1.97 (−0.88 to 4.82)

2.88 (1.17–4.59)***

Ever pregnant before college

‡

Parental connectedness

‡

−0.92 (−1.96 to 0.12)†

Parental attitude toward birth control

1.16 (0.46–1.86)***

0.47 (−0.07 to 1.02)†

Parental communication about sex

‡

Peer communication about sex

0.35 (−0.28 to 0.98)

†

Impulsivity

−1.09 (−2.26 to 0.09)†

−0.71 (−1.45 to 0.03)†

Sensation-seeking

‡

Conscientiousness

0.36 (−0.28 to 1.00)

0.31 (−0.17 to 0.79)

R2 within

0.18 (0.03–0.32)*

0.29 (0.15–0.43)***

R2 between

0.25 (0.15–0.35)***

0.20 (0.07–0.34)**

p<.05.

p<.01.

NIH-PA Author Manuscript

***

p<.001.

†

p<.10.

‡

Measure was constrained to 0 in the model.

Notes: Figures in parentheses are 95% confidence intervals. na=not applicable, because measure was not included in the model. ref=reference
group.

NIH-PA Author Manuscript
Perspect Sex Reprod Health. Author manuscript; available in PMC 2015 June 01.

File Type	application/pdf
File Modified	2018-01-19
File Created	2018-01-19