Mascarenhas et al.

Population Health Metrics 2012, 10:17

[Link]

RESEARCH Open Access

Measuring infertility in populations: constructing a

standard definition for use with demographic and
reproductive health surveys
Maya N Mascarenhas1, Hoiwan Cheung2, Colin D Mathers3 and Gretchen A Stevens3*

Abstract
Background: Infertility is a significant disability, yet there are no reliable estimates of its global prevalence. Studies
on infertility prevalence define the condition inconsistently, rendering the comparison of studies or quantitative
summaries of the literature difficult. This study analyzed key components of infertility to develop a definition that
can be consistently applied to globally available household survey data.
Methods: We proposed a standard definition of infertility and used it to generate prevalence estimates using 53
Demographic and Health Surveys (DHS). The analysis was restricted to the subset of DHS that contained detailed
fertility information collected through the reproductive health calendar. We performed sensitivity analyses for key
components of the definition and used these to inform our recommendations for each element of the definition.
Results: Exposure type (couple status, contraceptive use, and intent), exposure time, and outcomes were key
elements of the definition that we proposed. Our definition produced estimates that ranged from 0.6% to 3.4% for
primary infertility and 8.7% to 32.6% for secondary infertility. Our sensitivity analyses showed that using an exposure
measure of five years is less likely to misclassify fertile unions as infertile. Additionally, using a current, rather than
continuous, measure of contraceptive use over five years resulted in a median relative error in secondary infertility
of 20.7% (interquartile range of relative error [IQR]: 12.6%-26.9%), while not incorporating intent produced a
corresponding error in secondary infertility of 58.2% (IQR: 44.3%-67.9%).
Conclusions: In order to estimate the global burden of infertility, prevalence estimates using a consistent definition
need to be generated. Our analysis provided a recommended definition that could be applied to widely available
global household data. We also summarized potential biases that should be considered when making estimates of
infertility prevalence using household survey data.
Keywords: Infertility, Demographic and health surveys, Population health, Prevalence

Background global health community has paid considerable attention

Infertility can be a devastating condition for couples who to fertility, the importance of simultaneously addressing
want to have children. Infertility affects men and woman infertility cannot be discounted. Understanding the magni-
alike, as both genders report associated psychological tude and distribution of infertility is key to creating
distress, depression, and low self-esteem [1,2]. In many evidence-based policies and reinforcing efforts to reduce
cultures, the social repercussions of infertility compound the burden of this condition.
the individual impact. Infertility has been observed to re- A global picture of infertility is not available partly due
sult in divorce, loss of economic resources, and even the to the difficulty in defining the condition. In the litera-
annulment of rights to burial grounds [3]. Although the ture, infertility is used synonymously with sterility, infec-
undity, childlessness, and subfertility. These terms are
used both interchangeably and inconsistently; an explicit
* Correspondence: stevensg@[Link]
3
Department of Health Statistics and Informatics, World Health Organization,
detailing of each component of the definition is needed
Geneva, Switzerland to clarify what is being measured. Gurunath et al.
Full list of author information is available at the end of the article

© 2012 Mascarenhas et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the
Creative Commons Attribution License ([Link] which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is properly cited.
Mascarenhas et al. Population Health Metrics 2012, 10:17 Page 2 of 11
[Link]

highlighted this lack of consistency in definitions In this paper, we developed a demographic definition
through their systematic review of literature on preva- of infertility that may be applied to a range of publicly
lence studies measuring infertility [4]. The review available national household surveys for the purpose of
detailed the inconsistent specification and incomplete generating a consistent measure of the prevalence of
documentation of each of four components of the infer- infertility. All infertility measurements – clinical, epide-
tility definition – age range, exposure type, exposure miologic, and demographic – incorporate some time-
time, and outcome. dependent measurement of exposure, i.e., a couple is
This inconsistency partly stems from the variety of disci- classified as infertile if they have tried unsuccessfully to
plines that generate infertility measurements. As shown in become pregnant or give birth for more than a mini-
Table 1, the infertility definitions released by institutions mum length of time. Most demographic household sur-
that set guidelines for researchers have not agreed upon a veys do not directly collect information on the length
standard definition. The largest disparity lies between the of time that a couple has been trying to conceive.
clinical and the demographic definitions. The clinical defi- However, exposure can be inferred from a woman’s
nitions are oriented toward the early detection of infertility couple status, contraceptive use, and desire for a child,
in individual patients with the aim of starting treatment, if all measured over a defined period of time. These in-
necessary, as early as possible. The demographic definition, direct measures are available across most demographic
on the other hand, attempts to measure infertility on a and reproductive health surveys. The level of detail
population level, relying on widely applied household sur- available for each measure can vary by survey and may
veys rather than sparse data from clinical visits. The clin- lead to biases. In order to account for potential sources
ical definition is important for understanding infertility on of measurement error in our surveys, we evaluated the
an individual level, while the population measures pro- effects of varying individual components of our infertil-
duced by the demographic definition are important inputs ity definition and made recommendations accordingly.
to understanding the magnitude, distribution, and under- Previous studies have evaluated demographic defini-
lying trends of infertility at a population level. tions of infertility using simulation analyses, with the
Table 1 Definitions of infertility found in the literature
Reference Definition
International Committee for Monitoring Technology and World Infertility (clinical definition) is a disease of the reproductive system defined
Health Organization, 2009 Revised Glossary on ART by the failure to achieve a clinical pregnancy after 12 or more months of regular
Terminology [15] unprotected sexual intercourse.
American Society for Reproductive Medicine, 2008 Definitions Infertility is a disease, defined by the failure to achieve a successful pregnancy
of infertility and recurrent pregnancy loss [16] after 12 months or more of regular unprotected intercourse. Earlier evaluation
and treatment may be justified based on medical history and physical findings
and is warranted after six months for women over age 35 years.
National Institute for Health and Clinical Excellence Infertility should be defined as the failure to conceive after regular unprotected
guideline 2004 [17] sexual intercourse after two years in the absence of a known reproductive
pathology.
World Health Organization, 2001 Reproductive Health Percentage of women of reproductive age (15–49) at risk of pregnancy
Indicators for Global Monitoring [9] (not pregnant, sexually active, noncontracepting, and nonlactating) who
report trying for a pregnancy for two years or more.
World Health Organization, 1985 Manual for the Infertility, primary: The woman has never conceived despite cohabitation,
investigation and diagnosis of the infertile couple [18] exposure to pregnancy, and the wish to become pregnant for at least
12 months. Infertility, secondary: The woman has previously conceived but is
subsequently unable to conceive despite cohabitation, exposure to pregnancy,
and the wish to become pregnant for at least 12 months. If the woman has
breastfed a previous infant, then exposure to pregnancy should be calculated
from the onset of regular menstruation following delivery.
Demographic definition, 1985 The dictionary of The inability to produce a live birth. The term usually refers to women, but men
demography [19] or couples can be the focus of attention. Used without qualification, sterility
implies irreversibility, but the term temporary sterility is sometimes used.
A distinction is made between primary sterility where a woman has never been
able to have a child, and secondary sterility, which occurs after the birth of at
least one offspring.
World Health Organization, 1975 The Epidemiology Primary infertility: The woman has never conceived despite cohabitation and
of Infertility – Report of a WHO Scientific Group [8] exposure to pregnancy for at least two years. Secondary infertility: The woman
has previously conceived but is subsequently unable to conceive despite
cohabitation and exposure to pregnancy for a period of two years; if the woman
has breastfed a previous infant, then exposure to pregnancy should be calculated
from the end of the period of lactational amenorrhea.
Mascarenhas et al. Population Health Metrics 2012, 10:17 Page 3 of 11
[Link]

aim of constructing a definition for high-fertility set- obtaining comparable prevalence estimates. We propose a
tings [5,6]. Larsen and Menkin [5] used simulation ana- standard definition of primary and secondary infertility that
lyses to test the sensitivity of different measures of can be applied to these surveys, as detailed below:
infertility to survey sample size, age distribution of ster-
ility, age at marriage, and fecundity. They found that 1) Primary infertility is defined as the absence of a
the five-year “subsequently infertile” measure, defined live birth for couples that have been in a union for
as the percent of continuously married couples who do at least five years, during which neither partner
not bear a child during a five-year observation period, used contraception, and where the female partner
was most robust to the factors considered in their sen- expresses a desire for a child. The prevalence of
sitivity analyses. Larsen [6] extended her previous work primary infertility is calculated as the number of
to analyze the error arising from contraceptive use woman in an infertile union divided by the
through microsimulations. She found that contraceptive combined number of women in fertile and infertile
use could be ignored if fewer than 6% of women were unions. Women in a fertile union have had at least
users. Larsen [7] collected primary data in Northern one live birth and have been in a union for at least
Tanzania to compare the World Health Organization’s five years at the time of the survey (Figure 1).
(WHO) epidemiological definition of infertility, trying 2) Secondary infertility is defined as the absence of a
without success to bear a child for two years or more live birth for couples that have been in a union for at
[8,9], to one derived from data commonly available in least five years since the female partner’s last live
demographic and fertility surveys. She found that the birth, during which neither partner used
five-year “subsequently infertile” measure was appropri- contraception, and where the female partner
ate only when it was limited to women who had expresses a desire for a future child. The
expressed a desire to have a child in addition to being prevalence of secondary infertility is calculated as the
continuously married with no successful birth during number of women in an infertile union divided by the
the exposure period. combined number of women in infertile and fertile
We based our definition of infertility on the “subse- unions. Women in a fertile union have had at least
quently infertile” estimator proposed by Larsen and col- one live birth in the past five years and, at the time of
leagues; however, we extended their previous work in the survey, have been in a union for at least five years
this paper for two reasons. First, the proportion of following their first birth (Figure 2).
women of reproductive age using contraception has
increased around the world, ranging from 24% in Africa We used DHS data to explore how varying key com-
to 43% to 80% in other regions [10]. This significant ponents of each definition affected prevalence estimates.
change necessitates the use of a definition that is robust The components we assessed were exposure time, ex-
to contraceptive use greater than 6%. Second, more posure type (couple status, contraceptive use, and desire
detailed exposure measurements are now available from for a child), and outcome. To determine the effect of
many nationally representative surveys. Specifically, the varying each component with respect to our proposed
Demographic Health Survey (DHS) has now collected definition, we assessed the associated median relative
information on couple status and contraceptive use over errors and the interquartile range of relative errors, mea-
a five-year period in a number of countries around the sures which are not unduly affected by surveys with very
world. small sample sizes.
Our proposed definition may be used to estimate Every standard DHS questionnaire collects data from
country, regional, and global prevalence of infertility and women of reproductive age on couple status, birth his-
its trends from the large pool of available demographic tory, contraceptive use, and desire to have a child. A
household surveys. It is relevant across diverse popula- subset of DHS include the reproductive health calendar,
tions with different rates of marriage, contraceptive use, a tool used to collect detailed information on couple sta-
and childbearing practices. By presenting a consistent tus, contraceptive use, and birth history for each month
and widely applicable definition, as well as quantifying spanning the five years prior to the survey. In order to
the biases that arise from using common variations of compare the prevalence estimates generated using the
this definition, this paper provides a valuable framework detailed measures from the reproductive health calendar
for future estimates of the global burden of infertility. to those generated using standard survey questions, we
restricted our analyses to the subset of surveys that
Methods included both the standard questionnaire and the repro-
Few papers in the literature present estimates of infertility ductive health calendar.
prevalence using clear and complete definitions. Household Infertility measures like the one proposed here reflect
surveys provide a robust alternative to published studies for etiologies that can be attributed to the male partner, the
Mascarenhas et al. Population Health Metrics 2012, 10:17 Page 4 of 11
[Link]

Women
20-49 years

Not in a union
Union ≥5 years1
OR union <5 years

NOT EXPOSED /
No births ≥1 births CENSORED

No contraceptive use in
B- EXPOSED:
past 5 years AND
desires a child2 FERTILE

A - EXPOSED:
INFERTILE

Any contraceptive use in

past 5 years OR no
desire for a child2

NOT EXPOSED /
CENSORED

Figure 1 Primary infertility, women aged 20 to 49 years using a five-year exposure period. Primary infertility prevalence is calculated as
the number of infertile women (A) divided by the number of women who are both infertile and fertile (A and B). 1: Union is defined as marriage
or cohabitation. 2: Desire for a child is defined as wanting a child, undecided, or declared infecund.

female partner, or both partners. Because the DHS inter- Exposure time
view women of reproductive age, our analysis was Exposure time refers to the minimum time necessary to
indexed by the age of the women in each couple, and we assess whether a union is infertile. Using the reproduct-
refer to women and couples interchangeably in the text. ive health calendar, we produced an estimate of infertility
We excluded women who had been exposed for less that captured monthly couple status and contraceptive
than the minimum exposure period from the analysis. use over the exposure period. We compared the use of
Prevalence estimates were generated for six age groups one- and two-year exposure periods to the five-year ex-
(20–24, 25–29, 30–34, 35–39, 40–44, and 45–49) posure period used in our proposed definition.
indexed by the age of the female survey respondent at We found strong evidence of misclassification when
the time of the survey. We calculated prevalence using using exposure times of one and two years. Shorter ex-
sample weights to account for complex survey design. posure times are more sensitive to identifying women in
We generated age-standardized estimates using the an infertile union, while a longer exposure time is less
WHO reference population [11]. likely to misclassify women as infertile who, in the ab-
sence of intervention, would have given birth. The preva-
lence estimates for primary and secondary infertility
Results were inversely related to exposure time (Figure 3). The
We analyzed 53 surveys representing 26 countries and shorter exposure times of one and two years produced a
spanning the years 1990 to 2008 (Table 2). Using the strong age trend; prevalence estimates of infertility were
proposed definitions, we found that the age-standardized highest in the youngest age groups and declined as
prevalence of primary infertility ranged from 0.6% (95% women aged. This pattern of decreasing infertility preva-
confidence interval [CI]: 0.4%, 0.8%) in Peru in 1992, to lence in older women is highly implausible.
3.4% (95% CI: 2.7%, 4.1%) in Morocco in 1992. The age- Because the proposed definition uses birth as an out-
standardized prevalence of secondary infertility ranged come, the exposure period must accommodate gestation
from 8.7% (95% CI: 7.0%, 10.7%) in Zimbabwe in 2005 time in addition to time to conception. When shorter
to 32.6% (95% CI: 27.8%, 37.8%) in Armenia in 2000. exposure times are used, pregnant women are more
Mascarenhas et al. Population Health Metrics 2012, 10:17 Page 5 of 11
[Link]

Women
20-49 years

Union1 5 years No births OR not in a union

AND gave birth to first child OR in union <5 years OR gave
5 years ago birth to first child <5 years ago

NOT EXPOSED /
Last birth2 5 years Last birth <5 years
CENSORED

No contraceptive use in
B- EXPOSED:
past 5 years AND
desires a child3 FERTILE

A - EXPOSED:
INFERTILE

Any contraceptive use in

past 5 years OR no
desire for a child2

NOT EXPOSED /
CENSORED

Figure 2 Secondary infertility, women aged 20 to 49 years using a five-year exposure period. Secondary infertility prevalence is calculated
as the number of infertile women (A) divided by the number of women who are both infertile and fertile (A and B). 1: Union is defined as
marriage or cohabitation. 2: Last birth refers to the most recent birth after the first child. 3: Desire for a child is defined as wanting a child,
undecided, or declared infecund.

likely to be labeled as being in an infertile union because capture a continuous measure of union length for
the exposure period is less likely to be long enough to women who have been in multiple unions. Measuring
capture the birth. The percent of women who are classi- time since first union for women with multiple unions
fied as infertile but are pregnant at the time of the sur- has the potential to overestimate a woman's exposure
vey can be used as an indicator of misclassification. time to pregnancy and the corresponding prevalence of
Tellingly, this percentage decreases as exposure time infertility. Limiting the analysis to women in their first
increases for both primary and secondary infertility. For and only union, however, can produce an underestimate
primary infertility, the median age-standardized percent of the prevalence if infertile women are more likely to
of infertile women who are pregnant using a five-year have multiple unions or women with multiple unions
exposure time is 3.8% (interquartile range [IQR]: 1.7%- are predisposed to contract sexually transmitted diseases
5.3%) compared to 10.8% (IQR: 9.2%-12.6%) using a that cause infertility. In addition, DHS do not collect in-
one-year exposure time; for secondary infertility, the formation on past temporary separations, for example
percent increases from 3.9% (IQR: 1.8%-6.6%) to 10.2% due to migratory work, which can also result in an over-
(IQR: 7.6%-13.2%). estimate of the exposure time.
The reproductive health calendar collects detailed in-
Couple status formation on couple status during the five-year period
Marriage and cohabitation are measures found in all prior to the survey and can be used to assess the biases
standard household surveys. In many countries with from using the measurement of time from first, rather
high fertility rates, they also provide a good proxy for ex- than current, marriage. In order to assess the impact of
posure, particularly when combined with other variables incorporating time since first union into the exposure
assessing contraceptive use and intent. measure, we examined the effects of including women
In the majority of DHS, the duration of couple status with multiple unions and of limiting the sample to women
is measured from the time of first union (marriage or in their first and only union. We compared both sets of
cohabitation). As a result, it is impossible to accurately prevalence estimates to estimates produced by using the
Mascarenhas et al. Population Health Metrics 2012, 10:17 Page 6 of 11
[Link]

Table 2 Age-standardized primary and secondary infertility prevalence estimates, women aged 20-49 years as
calculated using selected Demographic and Health Surveys
Country Year Primary Secondary Survey size
Prevalence (%) Sample size Prevalence (%) Sample size
Armenia 2000 1.4 (1.0, 2.0) 3584 29.2 (24.8, 34.0) 745 5262
Armenia 2005 1.3 (0.9, 1.9) 3424 32.6 (27.8, 37.8) 647 5430
Bangladesh 1993 1.7 (1.4, 2.1) 7030 18.5 (16.4, 20.8) 3758 8225
Bangladesh 1996 1.5 (1.2, 1.9) 6497 21.8 (19.8, 24.0) 3240 7709
Bangladesh 1999 1.8 (1.5, 2.1) 7406 18.3 (16.3, 20.5) 3377 8922
Bangladesh 2004 1.7 (1.4, 2.0) 8114 20.5 (18.4, 22.8) 3610 9737
Bolivia 1994 1.0 (0.7, 1.4) 4099 9.5 (8.1, 11.2) 2525 6780
Brazil 1991 1.5 (1.1, 2.1) 2516 12.4 (9.4, 16.1) 1122 4805
Brazil 1996 1.6 (1.3, 2.1) 5668 16.7 (14.3, 19.4) 1825 10075
Colombia 1990 1.6 (1.1, 2.4) 3234 12.9 (9.0, 18.0) 1273 6835
Colombia 1995 1.4 (1.0, 1.9) 4326 13.2 (11.0, 15.8) 1655 8969
Colombia 2000 1.5 (1.1, 2.1) 4295 15.9 (13.2, 19.0) 1433 9319
Colombia 2005 1.5 (1.2, 1.8) 14159 15.9 (14.1, 17.9) 4591 31047
Democratic Republic of the Congo 2002 1.7 (1.4, 2.1) 10189 23.1 (20.6, 25.7) 3670 18576
Dominican Republic 1991 2.2 (1.6, 3.1) 2806 16.2 (12.1, 21.3) 1183 5596
Dominican Republic 1996 2.4 (1.8, 3.1) 3539 19.2 (15.9, 23.0) 1413 6584
Dominican Republic 1999 1.0 (0.4, 2.4) 532 23.7 (15.3, 34.8) 190 1018
Egypt 1995 2.8 (2.4, 3.2) 10817 13.2 (11.8, 14.8) 5666 14075
Egypt 2000 2.7 (2.3, 3.1) 11148 13.1 (11.7, 14.8) 5281 14978
Egypt 2005 2.6 (2.3, 3.0) 13719 15.0 (13.6, 16.6) 6255 18616
Egypt 2008 3.1 (2.7, 3.5) 11495 14.7 (13.2, 16.4) 4851 15891
Ethiopia 2005 1.3 (0.9, 1.7) 6617 11.7 (10.3, 13.1) 4883 10818
Guatemala 1995 1.0 (0.7, 1.4) 6384 16.6 (14.9, 18.4) 4379 9454
India 2005 2.7 (2.5, 2.9) 71095 24.6 (23.3, 26.0) 22740 100430
Indonesia 1991 2.6 (2.2, 3.1) 16961 18.0 (16.6, 19.6) 8289 21910
Indonesia 1994 2.1 (1.8, 2.5) 21059 21.2 (19.3, 23.3) 10080 27085
Indonesia 1997 2.6 (2.2, 3.0) 21295 20.4 (18.6, 22.3) 9509 27707
Indonesia 2002 2.6 (2.2, 3.0) 22087 18.8 (16.9, 20.8) 8847 28559
Indonesia 2007 2.1 (1.8, 2.5) 24616 17.7 (16.2, 19.3) 10479 31981
Jordan 1997 2.5 (2.0, 3.1) 4087 9.8 (8.1, 11.7) 2591 5342
Kazakhstan 1999 1.1 (0.7, 1.7) 2410 25.9 (20.5, 32.2) 609 4022
Kenya 1998 1.0 (0.7, 1.4) 3584 15.6 (13.3, 18.2) 2242 6029
Kenya 2003 1.0 (0.7, 1.5) 3508 11.6 (9.7, 13.8) 2241 6375
Malawi 2004 1.0 (0.7, 1.3) 5624 10.1 (8.8, 11.7) 4090 9291
Morocco 1992 3.4 (2.7, 4.1) 4059 10.8 (9.3, 12.5) 2568 7111
Nicaragua 1997 1.1 (0.9, 1.5) 6005 12.5 (10.7, 14.6) 3085 10277
Paraguay 1990 1.5 (1.2, 2.0) 2722 15.8 (13.8, 18.1) 1699 4533
Peru 1992 0.6 (0.4, 0.8) 7052 10.8 (9.5, 12.3) 3756 12398
Peru 1996 0.8 (0.6, 1.0) 13715 10.0 (8.9, 11.4) 7272 22897
Peru 2000 1.0 (0.8, 1.3) 12938 10.5 (9.3, 11.9) 5913 22095
Philippines 1993 2.1 (1.7, 2.5) 7190 15.1 (13.7, 16.7) 4102 11890
Philippines 1998 1.8 (1.5, 2.3) 6709 17.0 (15.3, 18.8) 3650 11034
Mascarenhas et al. Population Health Metrics 2012, 10:17 Page 7 of 11
[Link]

Table 2 Age-standardized primary and secondary infertility prevalence estimates, women aged 20-49 years as
calculated using selected Demographic and Health Surveys (Continued)
Philippines 2003 2.7 (2.1, 3.3) 6778 17.7 (16.0, 19.5) 3302 10987
Moldova 2005 3.2 (2.4, 4.3) 3894 29.5 (25.0, 34.4) 708 6037
Turkey 1993 2.3 (1.8, 2.8) 4916 18.0 (15.4, 21.0) 1688 6189
Turkey 1998 2.6 (2.1, 3.2) 4644 20.1 (17.5, 22.9) 1661 6813
Turkey 2003 2.3 (1.7, 3.1) 3322 15.9 (12.4, 20.1) 1084 7835
Tanzania 2004 1.8 (1.4, 2.3) 4714 16.5 (14.8, 18.3) 3511 8032
Viet Nam 1997 1.0 (0.7, 1.4) 4300 12.2 (9.5, 15.4) 1478 5551
Viet Nam 2002 0.8 (0.6, 1.2) 4504 13.8 (10.0, 18.7) 1059 5598
Zimbabwe 1994 1.1 (0.8, 1.7) 2650 14.6 (12.5, 17.1) 1813 4642
Zimbabwe 1999 1.6 (1.1, 2.2) 2385 13.9 (11.5, 16.8) 1472 4439
Zimbabwe 2005 1.2 (0.8, 1.6) 3567 8.7 (7.0, 10.7) 2050 6777
95% confidence intervals are shown in parentheses. Sample size refers to the proportion of women who were included in the calculations of primary or secondary
infertility (i.e., the denominator in the prevalence calculation), and the survey size refers to the total number of women surveyed.

reproductive health calendar, which captured a continuous period, we compared the prevalence estimates generated
measure of couple status over five years irrespective of from current use to those generated using a continuous
whether it was a first or subsequent union. The magnitude monthly measure.
of either sets of biases proved to be small. The median Prevalence estimates using a current, rather than con-
relative error in age-standardized prevalence from using tinuous, measure of contraceptive use produced esti-
time since first marriage for women in their first and only mates biased upwards for primary and secondary
union compared to using all women with a continuous infertility. However, the patterns were different for each
measure of couple status was −4.5% (IQR: -9.9%-0.0%) for type of infertility (Table 3). For primary infertility, the
primary infertility and −5.3% (IQR: -10.4% - -1.9%) for sec- median relative error between current and continuous
ondary infertility. The equivalent comparison of time since usage was largest in the younger age groups but was
first marriage and the continuous measure of couple sta- small for ages 30 and above (Table 3), producing median
tus amongst all women, including those with multiple age-standardized differences of 5.5% (IQR: 2.7%-14.9%).
marriages, produced median relative percent differences For secondary infertility, large differences were observed
of 5.2% (IQR: 0.5%-9.6%) for primary infertility and differ- across all age groups, with a median age-standardized
ences of 4.0% (IQR: 2.1%-9.7%) for secondary infertility. difference of 20.7% (IQR: 12.6%-26.9%).
Although information is not available on past separa-
tions during the exposure period, women are asked to
report whether their partners are currently living in the Intent
same household. We assessed the effect of excluding The inclusion of a woman’s intent to conceive serves
women whose husbands were living elsewhere at the as a proxy for unprotected sexual intercourse, be-
time of the survey from the infertility measure. We cause women who explicitly do not want children
found that the effect was negligible: the median relative are likely to employ protective measures to avoid
error in age-standardized prevalence was 0.6% (IQR: conception. Women who did not desire a child were
-2.0%-2.2%) for primary infertility and −1.2% (IQR: excluded from the infertile group, even if they met
-3.5% - -0.5%) for secondary infertility. all other criteria for infertility. Women who did not
provide an answer, were undecided, self-identified as
infertile, or stated that they wanted a child were
Contraceptive use assessed according to the remaining criteria for infer-
Unprotected sex is generally necessary for conception tility. Excluding women who did not desire a child
and thus serves as a useful proxy for exposure. The re- from the analysis did not notably affect primary in-
productive health calendar allowed us to measure the fertility estimates, especially in the younger ages
continuous use of contraception over the exposure (Table 3), producing a median age-standardized relative
period; however, most household surveys only collect in- difference of 4.1% (IQR: 1.7%-7.7%). For secondary in-
formation on current rather than continuous usage. In fertility, the prevalence differences were larger and pro-
order to determine whether current contraceptive use is duced an overall median age-standardized difference of
a sufficient proxy for continuous use over a five-year 58.2% (IQR: 44.3%-67.9%).
Mascarenhas et al. Population Health Metrics 2012, 10:17 Page 8 of 11
[Link]

Primary Infertility by Age and Exposure Time

15
Infertility Prevalence (%)
5 0 10

12 months 24 months 60 months

Age Groups
20-24 25-29 30-34 35-39 40-45 45-49

Secondary Infertility by Age and Exposure Time

100 80
Infertility Prevalence (%)
20 40 0 60

12 months 24 months 60 months

Age Groups
20-24 25-29 30-34 35-39 40-45 45-49

Figure 3 Boxplots of primary and secondary infertility prevalence by age for 12-, 24-, and 60-month exposure periods, 53
Demographic and Health Surveys. The boxplots depict the distribution of values for each estimate, showing the range with the whiskers and
the interquartile range and median with the box.

Outcome We used live birth as the outcome variable when

Because couples desire children, not simply pregnancies, assigning infertility to couples. As a result, some women
infertility affects couples regardless of whether the eti- who were pregnant at the time of survey were categor-
ology lies in conception or in the progression of the ized as being in an infertile union because they had not
pregnancy. Furthermore, distinguishing between the two yet achieved a live birth. This resulted in a potential
is often not possible in household surveys. Births are overestimate of prevalence. We calculated a set of preva-
reported quite accurately in household surveys, while lence estimates assuming that women were in a fertile
pregnancies that do not result in a birth (planned and union if they were pregnant at the time of survey. Not
spontaneous abortions) are not as reliably reported [12]. all women report they are pregnant during the first tri-
Spontaneous abortions are often underreported since mester so we expect that the proportion of self-reported
mothers may not be aware of their pregnancy status dur- pregnant women would not fully capture the potential
ing the first trimester when spontaneous abortions are misclassification; however, this effect is partially counter-
most likely to occur. Pregnancies that end in planned acted by pregnancies that do not result in births. For pri-
abortions are also underreported due to social stigma mary and secondary infertility, differences when using
and legal issues. pregnancy as the outcome were only notable for 20- to
Mascarenhas et al. Population Health Metrics 2012, 10:17 Page 9 of 11
[Link]

Table 3 Relative percent differences from sensitivity analyses of key components of the infertility
definition - contraception, intent, and outcome
Age group 20-24 25-29 30-34 35-39 40-44 45-49 Age-standardized
Primary infertility
Contraceptiona 9.2 5.8 4.3 0.0 0.0 0.0 5.5
(0.0, 31.9) (0.0, 19.2) (0.0, 13.2) (−0.1, 6.8) (−0.1, 3.5) (0.0, 0.0) (2.7, 14.9)
Intentb 0.0 0.0 0.0 2.2 2.1 11.4 4.1
(−0.1, 0.0) (0.0, 1.5) (0.0, 3.3) (0.0, 6.8) (0.0, 12.7) (0.0, 20.3) (1.7, 7.7)
Outcomec −10.1 −4.6 −1.5 0.0 0.0 0.0 −3.4
(−16.7, 0.0) (−9.1, 0.0) (−6.2, 0.0) (−1.4, 0.0) (0.0, 0.0) (0.0, 0.0) (−5.2, -1.7)
Secondary infertility
Contraceptiona 129.9 102.1 54.1 23.6 12.0 5.3 20.7
(73.1, 331.5) (57.6, 169.6) (32.4, 80.3) (12.4, 39.9) (6.9, 23.8) (2.4, 10.3) (12.6, 26.9)
Intentb 4.5 29.1 41.3 84.4 104.9 37.9 58.2
(0.0, 28.2) (16.8, 34.8) (27.0, 60.8) (59.6, 106.8) (65.9, 131.7) (20.8, 53.5) (44.3, 67.9)
Outcomec −16.7 −2.8 −0.8 0.0 0.0 0.0 −0.7
(−50.0, 0.0) (−5.9, 0.0) (−3.0, 0.0) (−0.5, 0.0) (0.0, 0.0) (0.0, 0.0) (−1.1, -0.3)
a- Current contraception (alternate) compared to continuous contraception (baseline) over the exposure period.
b- Not using the criteria of “desire for a child” variable (alternate) compared to using it to determine a couple's infertility status (baseline).
c- Assigning infertile woman who are pregnant as fertile (alternate) compared to classifying them as infertile (baseline).
Relative percent differences were calculated as the alternate prevalence less the baseline prevalence, divided by the baseline prevalence. The median and the
interquartile range across all surveys are shown.

29-year-olds (Table 3). This pattern may reflect the fact (Table 4). Our analysis produced a definition that can be
that woman in these age groups had the highest rates of applied consistently to a variety of household survey data
overall pregnancies. with information on fertility. We constructed this defin-
ition by considering the multiple components of infertility
Discussion and performing sensitivity analyses on each (Table 3). By
The literature lacks clear and consistent measures of infer- using information from the reproductive health calendar,
tility [4]. Through analyses of the DHS, we made recom- we were able to generate a definition that could assess
mendations for a definition of infertility with the goal of changing behaviors over a period of up to five years. We
informing future estimates of the burden of infertility compared these to prevalence estimates based on current

Table 4 Recommendations for defining infertility in analyses of household survey data based on sensitivity analyses of
key components of the infertility definition
Definitional Recommendation
component
Primary infertility Secondary infertility
Exposure time One- to two-year exposure periods increase misclassification
of fertile unions as infertile; a five-year exposure period
is recommended.
Couple status Measuring time since first union as a proxy for couple status
results in an acceptable error (< 5%) for prevalence estimates,
even for women with multiple unions. Temporary separations
have little effect on infertility estimates
Contraception Current contraception is a sufficient proxy for contraceptive Current contraception is not a sufficient proxy
use over the exposure period for women over 30. for contraceptive use over the exposure period.
Intent Intent has a small influence on prevalence estimates in the Disregarding intent increases estimates of infertility.
surveys analyzed, although this may not be true for Taking into account intent is recommended when
high-income settings. measuring the disability of secondary infertility.
Outcome Using reported birth is recommended as it is a more reliable
measure than reported pregnancies. When using birth as
an outcome, some women classified as infertile are pregnant
at the time of the survey. The proportion of infertile women
who are pregnant is smaller if longer exposure times are used.
Mascarenhas et al. Population Health Metrics 2012, 10:17 Page 10 of 11
[Link]

measures of behavior, which are more commonly found in Furthermore, survey questions built around sensitive
household surveys. topics such as childlessness and contraceptive use have the
We found that incorporating a five-year exposure time potential to produce responses biased by social norms.
was important for an accurate measure of infertility. We While household surveys are an invaluable source of data,
focused on improving specificity rather than sensitivity surveys that include explicit questions on infertility or con-
because demographic and health surveys tend to pro- tain detailed measures of important components are also
duce estimates of infertility that are biased upwards needed. Population-based studies that include clinical
when compared to more targeted infertility surveys assessments are a vital input to identifying the causes
[7,13]. Our analyses showed that an exposure measure underlying this condition and the relative contribution of
of five years, rather than one or two years, reduced the male and female factors.
likelihood of misclassifying fertile unions as infertile in
our estimates; a longer exposure period allows for the Conclusions
time that it takes to conceive and bear a child and pre- The definition presented in this paper can guide the use of
vents unreported temporary separations or periods of household survey data for the measurement of infertility
abstinence from unduly affecting the infertility measure. from sources such as the standard DHS, the World Fertil-
Prior analyses have also indicated that prevalence esti- ity Surveys, the Pan Arab Family and Child Health Sur-
mates generated using a five-year measure are most veys, the Centers for Disease Control Reproductive Health
similar to the prevalence of permanent sterility [5], Surveys, and the National Survey for Family and Growth
which causes the greatest health burden. Study (United States). Although the above surveys do not
Larsen has argued that couples who use contraceptives all collect detailed information on past reproductive health
are more fertile than those who do not, and thus should practices that were used in this paper, we have quantified
be classified as fertile [6,7]. However, it is by definition the potential biases and made corresponding recommen-
not possible to determine whether contraceptive users dations for how to best use the more commonly collected
would have conceived in the absence of contraceptive data found in these surveys (summarized in Table 4). The
use, regardless of the instrument used to determine in- analysis presented in this paper will allow for a compre-
fertility status. Worldwide, 63% of women of reproduct- hensive global understanding of infertility prevalence using
ive age currently use contraceptives [10]. Given the high the large number of nationally representative household
current prevalence of contraceptive use, it is likely that surveys available. Future analyses can build upon this work
users are more representative of the general population by examining the relationship of infertility, calculated
than they were in the high-fertility settings Larsen stud- using demographic and fertility surveys, with its determi-
ied. Thus, we argue that it is no longer possible to as- nants, including the prevalence of sexually transmitted dis-
sume that all couples using contraceptives are fertile, eases, female age at first pregnancy, and indicators of
and a robust definition of infertility should incorporate a nutritional status such as the prevalence of female anemia
measure of contraception over the exposure period. or underweight. A better understanding of the prevalence
There are limitations to using household surveys for esti- and causes of infertility, in turn, will enable informed pol-
mating the prevalence of infertility. Due to the nonspecific icy changes regarding prevention and treatment to effect-
nature of household survey questions, we had to construct ively reduce this global burden.
a definition of infertility that used couple status, contracep-
tive use, and intent to conceive as proxies for regular un- Abbreviations
DHS: Demographic and health surveys; IQR: Interquartile range; WHO: World
protected sexual intercourse, typically not measured in health organization.
household surveys. Assigning infertility to women who
expressed intent to conceive served as a proxy for regular, Competing interests
The authors declare that they have no competing interests.
unprotected sexual intercourse, and may correct for
underreporting of contraceptive use [7]. Nevertheless, Authors’ contributions
there is a risk that women who are unable to have a child MNM, GAS, and CDM developed the study content and design. MNM
extracted and analyzed survey data. MNM wrote the first draft of the report.
may cope by changing their stated or experienced fertility MNM, GAS, HC, and CDM all edited and approved the final version. GAS
preferences, which would discount the prevalence of infer- oversaw the research process. All authors read and approved the final
tility and underestimate its burden [14]. We extended the manuscript.
exposure period to reduce the likelihood of misclassifying
Acknowledgments
fertile couples as infertile. A longer exposure period, how- We thank Ties Boerma, Ulla Larsen, Sheryl Vanderpoel, and two reviewers for
ever, resulted in a smaller sample and increased the likeli- valuable comments and suggestions on the study design and the
hood of recall bias. There is also a risk that infertility can manuscript. This work was undertaken as a part of the Global Burden of
Diseases, Injuries, and Risk Factors 2010 Study. The results in this paper are
lead to voluntary dissolution of the union in fewer than prepared independently of the final estimates of the Global Burden of
five years, which would not be captured by our definition. Diseases, Injuries, and Risk Factors 2010 Study. A grant from the Bill &
Mascarenhas et al. Population Health Metrics 2012, 10:17 Page 11 of 11
[Link]

Melinda Gates Foundation supported the study’s core activities and partially
supported this study.

Author details
1
Department of Epidemiology and Biostatistics, University of California, San
Francisco, USA. 2The Geisel School of Medicine, Dartmouth University,
Hanover, USA. 3Department of Health Statistics and Informatics, World Health
Organization, Geneva, Switzerland.

Received: 18 January 2012 Accepted: 27 August 2012

Published: 31 August 2012

References
1. Chachamovich JR, Chachamovich E, Ezer H, Fleck MP, Knauth D, Passos EP:
Investigating quality of life and health-related quality of life in infertility:
a systematic review. J Psychosom Obstet Gynaecol 2010, 31:101–110.
2. Cui W: Mother or nothing: the agony of infertility. Bull World Health Organ
2010, 88:881–882.
3. Greil AL, Slauson-Blevins K, McQuillan J: The experience of infertility: a
review of recent literature. Sociol Health Illn 2010, 32:140–162.
4. Gurunath S, Pandian Z, Anderson RA, Bhattacharya S: Defining infertility–a
systematic review of prevalence studies. Hum Reprod Update 2011,
17:575–588.
5. Larsen U, Menken J: Measuring sterility from incomplete birth histories.
Demography 1989, 26:185–201.
6. Larsen U: Sterility in sub-Saharan Africa. Popul Stud 1994, 48:459–474.
7. Larsen U: Research on infertility: which definition should we use?
Fertil Steril 2005, 83:846–852.
8. WHO Scientific Group on the Epidemiology of Infertility: The epidemiology
of infertility. In Technical Report Series. Geneva: World Health Organization;
1975.
9. World Health Organization: Reproductive health indicators for global
monitoring: Report of the second interagency meeting, 2001. Geneva: World
Health Organization; 2001.
10. United Nations Department of Economic and Social Affairs Population
Division: Update for the MDG database: Contraceptive prevalence. New York:
United Nations Department of Economic and Social Affairs; 2011.
11. Ahmad O, Boschi-Pinto C, Lopez A, Murray C, Lozano R, Inoue M: Age
standardization of rates: a new WHO standard, Volume 31. Geneva: World
Health Organization; 2001.
12. Wilcox AJ, Horney LF: Accuracy of spontaneous abortion recall. Am J
Epidemiol 1984, 120:727–733.
13. Marchbanks PA, Peterson HB, Rubin GL, Wingo PA: Research on infertility:
definition makes a difference. The Cancer and Steroid Hormone Study
Group. Am J Epidemiol 1989, 130:259–267.
14. Saloman J, Murray C: A conceptual framework for understanding
adaptation, coping and adjustment in health state valuations. In
Summary Measures of Population Health, Volume 11.4. Geneva: World Health
Organization; 2002.
15. Zegers-Hochschild F, Adamson GD, de Mouzon J, Ishihara O, Mansour R,
Nygren K, Sullivan E, van der Poel S: The International Committee for
Monitoring Assisted Reproductive Technology (ICMART) and the World
Health Organization (WHO) Revised Glossary on ART Terminology, 2009.
Hum Reprod 2009, 24:2683–2687.
16. American Society for Reproductive Medicine Practice Committee: Definitions
of infertility and recurrent pregnancy loss. Fertil Steril 2008, 89:1603.
17. National Collaborating Centre for Women’s and Children’s Health: Fertility:
assessment and treatment for people with fertility problems. London: RCOG
Press; 2004. Submit your next manuscript to BioMed Central
18. Rowe PJ, Comhaire FH, Hargreave TB, Mellows HJ: WHO manual for the and take full advantage of:
standardized investigation and diagnosis of the infertile couple. Cambridge:
Cambridge University Press; 1993. • Convenient online submission
19. Pressat R, Wilson C: The dictionary of demography. New York: Blackwell
• Thorough peer review
Science; 1985.
• No space constraints or color figure charges
doi:10.1186/1478-7954-10-17 • Immediate publication on acceptance
Cite this article as: Mascarenhas et al.: Measuring infertility in populations:
• Inclusion in PubMed, CAS, Scopus and Google Scholar
constructing a standard definition for use with demographic and
reproductive health surveys. Population Health Metrics 2012 10:17. • Research which is freely available for redistribution

Submit your manuscript at

[Link]/submit