Infertility, perceived certainty of pregnancy, and contraceptive use in Malawi

Abstract

Infertility and unintended pregnancy are dual burdens in Malawi, where 41% of pregnancies are unintended and approximately 20% of people report infertility. Although preventing unintended pregnancy has been a focus in public health, infertility has rarely been explored as a factor that may be associated with contraceptive use. Using cross-sectional survey data (2017–2018; N=749), we report on the prevalence of and sociodemographic characteristics associated with infertility and certainty of becoming pregnant among women in Malawi. We conducted multivariable logistic regressions examining the relationship between infertility, certainty of becoming pregnant, and contraceptive use. Approximately 16% of women experienced infertility and three-quarters (78%) were certain they could become pregnant within one year. Women who experienced infertility had lower odds of contraceptive use than women who did not (AOR:0.56; 95%CI:0.39–0.83). Women who said there was ‘no chance’ or they were ‘unlikely’ to become pregnant also had lower odds of contraceptive use compared to women who were certain they would become pregnant (AOR: 0.30; 95% CI: 0.10–0.92). Our findings indicate that experiences and perceptions surrounding fertility are associated with contraceptive use, underscoring their importance in understanding how people manage their fertility to reach their reproductive goals.

Introduction

Globally, 44% of pregnancies are unintended, a proportion that has remained relatively constant over the past three decades, despite investment in research and programs aimed at reducing unintended pregnancy via increasing contraceptive use (Bearak et al. 2018). At the same time, a significant proportion of couples experience infertility, or the inability to become pregnant after 1–2 years of trying.¹ Accurate estimates of infertility are plagued by varying definitions and measures. Measurement is hindered by challenges ascertaining aspects of infertility that are necessary for constructing a more valid measure, including sexual frequency, length of time without conceiving, specifying between primary infertility (i.e., the inability to conceive a first pregnancy) and secondary infertility (i.e., the inability to conceive a second or higher order pregnancy), and distinguishing between infertility at the individual or couple-level (i.e., male/female/both), among many others (Mascarenhas, Cheung, et al. 2012). Based on differences in definition and data collection, estimates of infertility prevalence vary widely. An oft-cited lifetime prevalence of infertility – combining both primary and secondary infertility and using a 12-month definition – is 15% of couples of reproductive age (Mascarenhas, Cheung, et al. 2012; Boivin et al. 2007; Inhorn 2009). However, other global estimates are considerably higher, with combined primary and secondary infertility estimated to be up to 25–30% of reproductive-age couples in some regions (Rutstein and Shah 2004; Mascarenhas, Flaxman, et al. 2012; Mascarenhas, Cheung, et al. 2012; Nachitgall 2006).

The dual burden of unintended pregnancy and infertility has rarely been explored within public health; however, emerging research suggests that one reason high rates of unintended pregnancy persist is because people are reluctant to use contraception to prevent an unintended pregnancy if they are not sure they will be able to achieve an intended pregnancy when they desire (Johnson et al. 2018). While preventing unintended pregnancies and achieving intended pregnancies are pillars of reproductive rights (UNFPA, 1994), the latter has been largely ignored in public health and reproductive health, receiving substantially less monetary and intellectual investment than efforts focused on preventing unintended pregnancy (Gipson, Bornstein, and Hindin 2020).

The present study explores the relationships between self-reported experienced infertility, perceptions of infertility (certainty of ability to become pregnant), and contraceptive use in Malawi. We hypothesize that low perceived chance of becoming pregnant – influenced by both past experiences with infertility (i.e., inability to conceive after two years of trying) and low certainty of one’s ability to become pregnant – lower the odds of contraceptive use (Figure 1). In Malawi, pronatalist norms exist alongside a growing desire for smaller families (National Statistics Office (NSO) [Malawi] and ICF, 2017) and relatively high rates of self-reported infertility (Rao et al. 2018; Barden-O’Fallon 2005; Polis et al. 2020). These joint realities create a dynamic and multifaceted context in which individuals make decisions about preventing and pursuing pregnancy. Yet, there remain significant gaps in our understanting of how people make contraceptive decisions with their current and future pregnancy desires in mind.

FIGURE 1.

Hypothesized mechanisms between infertility-related constructs and contraceptive use

This study has four aims: (1) to describe the prevalence of and characteristics of women who report experiencing infertility; (2) to describe the characteristics of women who report various levels of certainty of their ability to become pregnant; (3) to examine the associations between experienced infertility and contraceptive use; and, (4) to examine the association between certainty of ability to become pregnant and contraceptive use. Understanding these relationships may be critical to addressing persistently high rates of unintended pregnancy, both in Malawi and globally.

Infertility-related Constructs and Contraceptive Use

We hypothesize that perceived certainty of becoming pregnant in the future and past experiences of infertility contribute to one’s overall perceived chance of becoming pregnant and, in turn, contraceptive use (Figure 1). Although fear of infertility, or other perceived side effects from contraception, are not addressed in this study, there is a growing body of literature around fear of side effects (including infertility) as a reason for contraceptive non-use. We acknowledge that there may be a bidirectional relationship between contraceptive use and fear/belief that contraception may cause infertility (Figure 1). (Sedlander et al. 2018; Huber-Krum and Norris 2020; Boivin et al. 2020).

Perceived chance of becoming pregnant

Several theories of health behavior (e.g., Theory of Reasoned Action and the Health Belief Model) highlight the role of perceived risk of an outcome or ailment (e.g., risk of becoming pregnant when a pregnancy is not desired) as a necessary precursor to implementing a preventative behavior (e.g., contraceptive use) (Rosenstock 1974; Hall 2012). Thus, low perceived risk (what we refer to as chance) of pregnancy may be a factor that contributes to contraceptive non-use.

The relationship between low perceived chance of pregnancy and contraceptive non-use has been demonstrated in existing literature. The belief that one cannot, or is unlikely to, get pregnant is an oft-reported reason for contraceptive non-use among women who do not want to become pregnant. In two U.S. studies, 33–42% of pregnant women with undesired pregnancies were not using contraception at the time they became pregnant because they did not believe they could become pregnant (Foster et al. 2012; Nettleman et al. 2007). One study examining unmet need for contraception in Malawi found that the perception that one cannot become pregnant for biological reasons may account for more than a quarter (27%) of contraceptive non-use among women in need of contraception (Westoff 2012). We hypothesize that two potentially related constructs make up perceived chance of pregnancy: certainty of ability to become pregnant and past experience with infertility, that is the inability to become pregnant after 1–2 years of trying. These two related constructs represent both anticipation of future fertility and past experiences with fertility, which, along with other factors, may influence decisions around contraceptive use. Separately and together, we will test if certainty of ability to become pregnant and experiences with infertility are associated with contraceptive use.

Certainty of ability to become pregnant

Findings from a nascent body of literature indicate important and understudied linkages between perceived infertility, certainty of ability to become pregnant, and contraceptive use. However, the few studies that examine constructs related to certainty of becoming pregnant vary considerably, likely due to the lack of standardized and/or validated measures (Gemmill 2018; Polis and Zabin 2012; Biggs and Foster 2013; Foster et al. 2012). In one study (Gemmill, 2018), using a question from the National Longitudinal Survey of Youth (NLSY), “Suppose you started to have unprotected intercourse today. What is the percent chance you would have a child within the next two years?” (0–100%), investigators found that women who reported a lower chance (0–50%; 50–75%) were less likely to use contraception, as compared to women who reported a higher chance of having a child (75–100%) (Gemmill 2018).

Experiences with infertility

The relationship between experienced infertility (i.e., unsuccessfully achieving pregnancy after 1–2 years of trying) and contraceptive use is largely unknown. However, theories of health behavior support that past experiences influence future behavior (Rosenstock 1974; Hall 2012). Thus, experiencing infertility may reduce one’s perceived chance of pregnancy and their motivation to use contraception to prevent an unintended pregnancy.

There has also been remarkably little research on who reports infertility, particularly in low-resource settings where a clinical diagnosis of infertility and infertility treatment are historically and currently rare (Starrs et al. 2018; Ombelet 2014; Johnson et al. 2018). Even in higher-resource settings, studies on infertility often only include those who seek treatment, which is a select group and not representative of the broader population who experience infertility (Greil, Slauson-Blevins, and McQuillan 2010). Having data solely from these highly selective samples from relatively high-resource settings has limited our understanding of the magnitude of infertility, the characteristics of people who experience infertility, and their contraceptive behaviors.

Belief that contraception causes infertility

While not directly addressed in this study, the belief that contraception may be a cause of infertility is well-documented (Richards 2002; Chipeta, Chimwaza, and Kalilani-Phiri 2010; Sedlander et al. 2018; Bornstein, Huber-Krum, et al. 2020). A study in Malawi found that women’s perceptions that a contraceptive method would affect their future fertility was associated with contraceptive method preference (Huber-Krum and Norris 2020). Another study found that perceiving that a specific contraceptive method did not cause infertility was associated with intention to use that method (Mumah et al. 2018; Machiyama et al. 2018).

Context

In Malawi, fertility and infertility are experienced within the broader social context of marriage, divorce, and family formation. Relationships are solidified through childbearing. Divorce and remarriage are also common (Reniers 2003). The commonality of relationship changes and remarriage underscore the importance of maintaining fertility, even after one achieves their desired number of children with a particular partner (Reniers 2003). Should a divorce occur, one’s fertility regains importance to solidify a new relationship (Reniers 2003). Even within a single partnership, pregnancy intentions and desired family size can change frequently (Yeatman and Sennott 2015; Yeatman, Sennott, and Culpepper 2013; Gibby and Luke 2019).

The total fertility rate in Malawi, at 4.4 children per woman, exceeds the reported wanted fertility rate of 3.4 children per woman. Despite a steady decline in unmet need for contraception (19% of women as of 2015–2016), approximately 30% of pregnancies in Malawi are mistimed and an additional 11% are considered unwanted (NSO [Malawi] and ICF 2017). Malawi’s relatively high total fertility rate and the discrepancy between total fertility and wanted fertility obscures the fact that Malawi also has a high rate of self-reported infertility. Two studies in different areas of Malawi from 2005 and 2018 found that approximately 20% of women (Rao et al. 2018) and 20% of women and men (Barden-O’Fallon 2005) reported ever experiencing infertility or difficulty becoming pregnant. The former (Rao et al., 2018) measured self-reported infertility using a two-year definition within a cohort of 915 women ages 22–32 years in rural Malawi. The latter (Barden-O’Fallon, 2005) examined self-reported difficulty becoming pregnant among a population of 678 women and 362 men. A fifth of women (20%) self-identified as experiencing difficulty getting pregnant. Among these women, 38% considered themselves or their partner to be infertile.

Sub-Saharan Africa is widely considered to have one of the highest rates of infertility globally (Mascarenhas, Flaxman, et al. 2012), yet several studies have shown that the general population often has misconceptions about the causes of infertility (Richards 2002; Sedlander et al. 2018; Bornstein, Huber-Krum, et al. 2020; Chipeta, Chimwaza, and Kalilani-Phiri 2010). Moreover, treatment is rarely available within the formal healthcare system and people commonly rely on traditional interventions (Barden-O’Fallon 2005; Parrott 2014). Given the concurrently high rates of unintended pregnancy (41%) (NSO [Malawi] and ICF 2017), and self-reported infertility (20%) in Malawi (Rao et al. 2018), it is imperative to assess the extent to which reproductive decisions – specifically contraceptive use – may be affected by one’s experience and perceptions of their fertility.

Methods

Study design

Data for this study come from the Umoyo Wa Thanzi (UTHA; Health for Life) research program, a cohort study focused on sexual and reproductive health decision making among women in Central Malawi. The cohort was recruited from the catchment area (approximately 20,000 residents) of a rural, non-profit hospital in 2013. In total, 68 villages were collapsed into 43 clusters based on size and geographic proximity. The clusters were then stratified into three geographic groups: rural, plantation, and trading center. Within these three categories, 11 clusters (19 villages) were randomly selected for inclusion in the cohort, resulting in eight rural, one plantation, and two trading center clusters (Huber et al. 2017). All women ages 15–49 residing in households within the selected villages were eligible to participate. In total, 1,034 women participated in Wave 1 (2014), representing a 96% response rate (Rao et al. 2018). Subsequent data collection occurred in Wave 2 (2016), Wave 3 (2017), and Wave 4 (2017–2018). There was a 75% retention rate between Waves 1 and 4 and an 85% retention rate between Waves 3 and 4. For Wave 4, enrollment efforts were expanded such that women living within the selected villages and who met the inclusion criteria, but were not previously enrolled in the study, were invited to participate. Data for this analysis are from Wave 4, where the total sample size was 1,161.

At each data collection point, the surveys were developed in English and then translated into Chichewa through an iterative process with both English-speaking and bilingual English-Chichewa team members. Translations were reviewed for meaning, with the final wording determined through collaborative consensus (Colina et al. 2016). Participants gave verbal and written consent to participate and re-consented at each wave. At Wave 4, participants were compensated with 2,000 MK (approximately $1.50–2.00 U.S. dollars). This study was approved by the Institutional Review Boards (IRB) at The Ohio State University, the Malawi College of Medicine, and the University of California, Los Angeles.

Analytic sample

Of the total 1,161 women who participated in Wave 4 of the UTHA study, the analytic sample excludes women who had never had sex (n=44), were sterilized (n=127), had post-partum amenorrhea or were post-menopausal (n=10), or were pregnant at the time of data collection (n=106) (total n=874). We additionally excluded women for whom we did not have full information available, resulting in a total of 749 women. The majority of women excluded due to missing information did not report age (n=45) or whether or not they experienced infertility (n=49).

There were some statistical differences between the analytic and excluded population (data not shown). About half of the 80 women who had age data but were excluded based on other missing data were between the ages of 15–19, compared to about 7% of the 749 women included in the analysis. Excluded women were more likely to be unmarried (61% of excluded vs. 15% of included women), and just under half of excluded women reported having sex in the previous three months, compared to 80% of women in the analytic sample. Excluded women had lower rates of contraceptive use compared to women in the analytic sample (62% vs. 77%) and were also less likely to report that they were certain to become pregnant within a year (63% vs. 78%). There were no differences between the analytic and excluded sample in terms of desire for a(nother) child or experienced infertility.

Variables

Independent variables:

Self-reported infertility:

Self-reported infertility was constructed from the survey question, “Have you ever tried to conceive a pregnancy for two years or longer without conceiving in that time?” Two years is a definition of infertility commonly used in epidemiological studies (Mascarenhas, Cheung, et al. 2012). Response options were yes, no, and never tried to conceive. The full distribution is included in Table 1. The majority of participants selected yes or no and those who responded ‘never tried to conceive’ were excluded in multivariable analyses.

TABLE 1.

Participant characteristics (N=749)¹

	Mean
Number of living children 2	2.79
Number of years of education	5.3
Age	28.1
	%
Age group
15–19	6.8%
20–24	29.0%
25–29	26.0%
30–34	20.0%
35–39	12.3%
40–45	5.9%
Currently married/cohabiting
Yes - monogamous relationship	68.0%
Yes - polygamous relationship	17.2%
No - not currently married/cohabiting	12.6%
No - never married/cohabited	2.3%
Ever been divorced (among ever-married)
Yes	31.2%
No	68.9%
Ever wants a(nother) child
Yes	76.8%
No	23.2%
Had sex in previous 3 months
Yes	80.0%
No	20.0%
Current contraceptive use
Yes	77.4%
No	22.6%
Certainty of pregnancy 3
Certain	78.0%
Likely	16.7%
No chance/ unlikely	5.3%
Experienced infertility (a period of 2+ years trying to achieve pregnancy) 4
Yes	16.2%
No	83.8%

¹Columns may not total to 100% due to rounding

²21 women did not report # of living children; thus, N=728 for this variable in Tables 1–4

³If you were to have sex and not use any method of contraception, how likely is it that you would become pregnant in the next year?

⁴Have you ever tried to conceive a pregnancy for two years or longer without conceiving in that time?

Certainty of pregnancy:

There are no validated measures of certainty of ability to become pregnant. Because of this, we used a novel measure developed for the UTHA study. We measured this construct using the following survey question: “If you were to have sex and not use any method of contraception, how likely is it that you would become pregnant in the next year?” There were five response options: no chance, unlikely, likely, certain, and don’t know. A one year timeframe was used because it is a typical medical definition of infertility (Mascarenhas, Cheung, et al. 2012), and, in Malawi, is also commonly considered to be the greatest length of time that it should take for conception to occur before people assume there is a problem (Bornstein, Gipson, et al. 2020). Due to the highly skewed distribution of this variable and precedence from the use of a similar measure in Malawi (Polis et al. 2020), we recoded it as a trichotomous variable (certain vs. likely vs. no chance/unlikely).

Dependent variable:

In the multivariable analyses, we examined current contraceptive use among women, assessed using the question, “Currently, are you using any method to avoid pregnancy in your relationship, whether it is a traditional or modern method?” If participants responded yes to this question, they were asked what method(s) they were using and less than 1% of women reported using a traditional method. Contraceptive use was coded as yes/no.

Covariates:

We examined sociodemographic variables that are often associated with contraceptive use, including marital/cohabitation status as a categorical variable (current monogamous relationship, current polygamous relationship, not currently married/cohabiting, and never married/cohabited), ever divorced (yes/no), age (categorical in 5-year increments), sexual activity in the previous three months (yes/no), years of education (continuous), number of living children (continuous), and desire for a(nother) pregnancy (yes/no) (Mandiwa et al. 2018; Digitale et al. 2017; Huber et al. 2017). Covariates included in the multivariable models were selected based on significant bivariate relationships, theoretical associations with contraceptive use, and model selection techniques to minimize multicollinearity, including correlation matrices and assessing Variance Inflation Factor (VIF).

Analyses

We first looked at the characteristics of women who reported infertility (Table 2) and certainty of becoming pregnant within one year of having sex without using any method of contraception (Table 3). We used chi² tests of independence and F-tests to examine differences in means. After examining the characteristics of individuals who reported infertility and certainty of becoming pregnant, we assessed the association between each variable and current contraceptive use, also using chi² tests of independence and F-tests (Table 4). Finally, we constructed three multivariable logistic regression models examining the relationship between infertility and contraceptive use; certainty of becoming pregnant and contraceptive use; and a third model that examined contraceptive use controlling for both infertility and certainty (Table 5).

TABLE 2.

Relationships between sociodemographic characteristics and self-reported infertility (N=749)¹

	No infertility	Experienced infertility	p-value
Total	83.8%	16.2%
Number of living children (mean)	2.75	2.93	0.292
Number of years of education (mean)	5.35	5.17	0.559
Age (mean)	27.7	29.8	0.0013**
Age group			0.026*
15–19	7.3%	4.1%
20–24	30.3%	22.3%
25–29	26.4%	24.0%
30–34	18.8%	26.5%
35–39	12.3%	12.4%
40–45	4.9%	10.7%
Currently married/cohabiting			0.124
Current monogamous relationship	68.6%	64.5%
Current polygamous relationship	16.2%	22.3%
Not currently married nor cohabiting	12.4%	13.2%
Never married nor cohabited	2.7%	0.0%
Ever been divorced (among ever-married)	29.3%	50.5%	0.015*
Ever wants a(nother) child	77.7%	71.9%	0.166
Had sex in previous 3 months	80.7%	76.0%	0.237
Certainty of pregnancy			0.000***
Certain	78.5%	75.2%
Likely	17.8%	10.7%
No chance/unlikely	3.7%	14.1%

¹p-values are for chi-square tests (categorical) and F-tests (continuous); columns may not total to 100% due to rounding; *p<0.05, **p<0.01. ***p<0.001

TABLE 3.

Relationships between sociodemographic characteristics and certainty of pregnancy after one year without using contraception (N=749)¹

	No chance/unlikely	Likely	Certain	p-value
Total	5.3%	16.7%	78.0%
Number of living children (mean)	3.00	3.09	2.7	0.042*
Number of years of education (mean)	5.23	5.49	5.29	0.806
Age (mean)	28.3	28.5	27.9	0.632
Age group				0.016*
15–19	5.0%	4.0%	7.5%
20–24	40.0%	31.2%	27.7%
25–29	25.0%	24.0%	26.5%
30–34	5.0%	20.0%	21.1%
35–39	7.5%	12.8%	12.5%
40–45	17.5%	8.0%	4.6%
Currently married/cohabiting				0.000***
Current monogamous relationship	47.5%	68.8%	69.2%
Current polygamous relationship	10.0%	20.0%	17.1%
Not currently married nor cohabiting	32.5%	9.6%	11.8%
Never married nor cohabited	10.0%	1.6%	1.9%
Ever been divorced (among ever-married)	41.7%	28.5%	31.1%	0.321
Ever wants a(nother) child	70.0%	78.4%	76.9%	0.544
Had sex in previous 3 months	67.5%	80.8%	80.7%	0.128
Self-reported infertility	42.5%	10.4%	15.6%	0.000***

¹p-values are for chi-square tests (categorical) and F-tests (continuous); columns may not total to 100% due to rounding; *p<0.05, **p<0.01. ***p<0.001

TABLE 4.

Relationship between independent variables and contraceptive use among women (N=749)^1,2

	Currently using contraception
	No	Yes	p-value
Total	22.6%	77.4%
Number of living children (mean)	2.85	2.76	0.52
Number of years of education (mean)	4.8	5.47	0.015*
Self-reported infertility			0.005**
Yes	32.2%	67.8%
No	20.7%	79.3%
Certainty of pregnancy			0.000***
Certain	20.2%	79.8%
Likely	24.0%	76.0%
No chance/ unlikely	52.5%	47.5%
Age (mean)	28.8	27.8	0.086
Age group			0.000***
15–19	47.1%	52.9%
20–24	16.1%	83.9%
25–29	19.5%	80.5%
30–34	19.3%	80.7%
35–39	23.9%	76.1%
40–45	47.7%	52.3%
Currently married/cohabiting			0.000***
Current monogamous relationship	15.7%	84.3%
Current polygamous relationship	17.0%	83.0%
Not currently married/cohabiting4	60.4%	39.6%
Ever wants a(nother) child			0.571
Yes	22.1%	77.9%
No	24.1%	75.9%
Had sex in previous 3 months			0.000***
Yes	14.0%	86.0%
No	56.7%	43.3%

¹p-values are for chi-square tests (categorical) and F-tests (continuous); rows total to 100%; *p<0.05, **p<0.01. ***p<0.001

²N is the women’s analytic sample

³Category combines not currently married and never married women

⁴Category combines ‘not currently’ and ‘never’ married or cohabiting

TABLE 5.

Bivariable and multivariable logistic associations between infertility variables, sociodemographic characteristics and current contraceptive use (outcome) among women (N=749)¹

	Bivariable		Model 1		Model2		Model 3
	OR	95% CI	AOR	95% CI	AOR	95% CI	AOR	95% CI
Self-reported infertility	0.54**	0.37–0.82	0.51***	0.35–0.74	- -	- -	0.56**	0.39–0.83
Certainty of pregnancy			- -	- -
Certain	1				1		1
Likely	0.80	0.48–1.34			0.73	0.45–1.17	0.70	0.43–1.13
No chance/unlikely	0.23***	0.11–0.50			0.26*	0.09–0.72	0.30*	0.10–0.92
Age group
15–19	0.21***	0.13–0.37	0.31**	0.15–0.66	0.28**	0.13–0.61	0.28**	0.13–0.61
20–24 (ref)	1		1		1		1
25–29	0.79	0.50–1.27	0.65	0.38–1.13	0.61	0.36–1.04	0.62	0.36–1.07
30–34	0.80	0.47–1.38	0.71	0.41–1.23	0.59*	0.35–0.99	0.63	0.37–1.08
35–39	0.61	0.30–1.24	0.55	0.28–1.06	0.49*	0.26–0.92	0.50*	0.27–0.94
40–45	0.21***	0.12–0.38	0.18***	0.09–0.37	0.17***	0.09–0.35	0.19***	0.10–0.37
Currently married/cohabiting
Current monogamous relationship	1		1		1		1
Current polygamous relationship	0.91	0.52–1.57	1.12	0.67–1.86	1.09	0.66–1.79	1.12	0.67–1.88
Not currently married/cohabiting2	0.12***	0.08–0.19	0.34***	0.20–0.56	0.41**	0.23–0.75	0.38**	0.21–0.71
Had sex in previous 3 months	8.02***	5.38–11.95	4.67***	2.59–8.40	5.19***	2.84–9.46	5.02***	2.64–9.53
Ever want a(nother) child	1.12	0.70–1.79	0.52*	0.28–0.96	0.51*	0.28–0.94	0.51*	0.28–0.95
Years of education	1.07**	1.02–1.12	1.07**	1.02–1.12	1.07**	1.02–1.12	1.07**	1.02–1.12

¹*p<0.05; **p<0.01; ***p<0.001

²Category combines ‘not currently’ and ‘never’ married or cohabiting

Model selection:

We refined the multivariable models examining contraceptive use by assessing correlations and multicollinearity between independent variables using a correlation matrix prior to inclusion. Age and number of living children were highly correlated (r=0.76; p<0.001) (not shown). We omitted number of living children from the multivariable models because it was not associated with contraceptive use among women in an F-test (Table 4). We also excluded the ‘ever divorced’ indicator variable in multivariable models because of its relationship with current marital/cohabiting status. In our final multivariable models (Table 5), we examined Variance Inflation Factor (VIF) values to assess multicollinearity of our independent variables. VIF values ranged from 1.13–5.38 (not shown). Acceptable VIF values are typically considered <10, and because relatively high VIF values were only observed in control variables, and not central to our interpretations, a slightly higher VIF is considered an insufficient reason to exclude the variable (O’Brien 2016; Allison 2012).

We conducted a series of sensitivity analyses to assess the robustness of our findings. First, we included number of living children instead of age in the multivariable models; however, results did not vary substantially for any of the independent variables (not shown). We also included alternative configurations of the desire for more children variable (now/in the next two years, after two years, undecided timing, never) and there were no substantial changes in the magnitude, direction, or significance of findings (not shown).

We conducted two additional sensitivity analyses to look at sub-groups within our sample. First, we looked at the multivariable models removing women who wanted to become pregnant now. Results did not change in significance or magnitude. Ultimately, we kept women who wanted to become pregnant now in the multivariable models (N=35), as 37% of them were currently using contraception (not shown). This may reflect challenges in measuring pregnancy desires: someone may want to become pregnant now, but act in accordance with other realities of their lives that lead them to prevent pregnancy (Santelli et al. 2003).

Ultimately, we constructed three logistic regression models examining contraceptive use as the outcome. The first model focused on the relationship between experienced infertility and contraceptive use; the second model focused on certainty of becoming pregnant and contraceptive use; and the third model included both main independent variables.

Results

Participant characteristics

Most (85%) of women in the sample were married/cohabiting, whether in a monogamous or polygamous relationship. About a third had been divorced at least once (31%), (Table 1). On average, women reported 2.8 children and 77% wanted a(nother) child sometime in the future. Three-quarters were currently using a method of contraception (77%) and approximately the same proportion were certain that they would become pregnant if they had sex and did not use a method of contraception for one-year (78%). Approximately 16% of women reported experiencing infertility or having tried unsuccessfully to become pregnant for a period of at least two years.

Correlates of infertility and certainty of pregnancy measures

Self-reported infertility

We next examine relationships between sociodemographic factors and self-reported infertility (Table 2). Women who reported that they experienced infertility were older (30 vs. 28; p<0.01) and more likely to have been divorced 51% vs. 29%; p<0.05) than women who did not report infertility (Table 2). Women who reported infertility were more likely to report that there was ‘no chance’ or it was ‘unlikely’ that they could become pregnant (14% vs. 4%) and less likely to report that they were ‘likely’ to become pregnant (11% vs. 18%), yet a similar proportion of women who did and did not experience infertility expressed that they were ‘certain’ they could become pregnant (75% vs. 79%) (p<0.001) (Table 2).

Certainty of pregnancy

Next, we examined the relationships between sociodemographic factors and certainty of pregnancy within a year of not using contraception (no chance/unlikely, likely, certain) (Table 3). A smaller proportion of women who were married in a monogamous relationship said ‘no chance/unlikely’ compared to ‘likely’ or ‘certain’ (48% vs. 69% for both likely and certain; p<0.001) (Table 3). Among women who said ‘no chance/unlikely,’ 43% reported infertility, while 10% of women who said pregnancy was ‘likely’ experienced infertility and 16% of women who said pregnancy was certain reported infertility (p<0.001). Women who were ‘certain’ of pregnancy had the least number of children (2.7 vs. 3.0–3.1; p<0.05).

Infertility measures and contraceptive use

Both self-reported experienced infertility and certainty of pregnancy were significantly associated with contraceptive use among women (Table 4). Women who experienced infertility were less likely to use contraception than those who had not experienced infertility (68% vs. 79%; p<0.01). Women who said they were ‘certain’ or ‘likely’ to become pregnant within a year of not using contraception were more likely to be using contraception than women who said there was ‘no chance’ or said it was ‘unlikely’ they would become pregnant (80% contraceptive use among women who were certain, 76% among likely, and 48% for no chance/unlikely) (p<0.001) Women who were married/cohabiting in monogamous or polygamous partnerships also had higher rates of contraceptive use than women who were not married/cohabiting (84% of monogamous, 83% of polygamous, and 40% of not married/cohabiting were using contraception) (p<0.001). Sex in the previous three months was also associated with contraceptive use, such that 86% of women who had sex in the past three months were using contraception compared to 43% of women who had not had sex in the previous three months (p<0.001) (Table 4).

Multivariable analysis

Table 5 reports on the bivariable relationships between all independent variables and contraceptive use, as well as three multivariable models examining: self-reported infertility and contraceptive use (Model 1), certainty of pregnancy and contraceptive use (Model 2), and a third multivariable model examining contraceptive use with both infertility and certainty of pregnancy as independent variables (Model 3).

Reported infertility and contraceptive use (Model 1):

Reporting infertility was associated with significantly lower odds of contraceptive use (AOR: 0.51; 95% CI: 0.35–0.74). Additionally, women who wanted a(nother) child had lower odds of contraceptive use compared to women who did not want more children (AOR: 0.52; 95% CI: 0.28–0.96), as did women who were not married/cohabiting compared to women in monogamous relationships (AOR: 0.34; 95% CI: 0.20–0.56). Education was significant, such that with every additional year of education, women had 7% higher odds of using contraception (AOR: 1.07; 95% CI: 1.02–1.12). The youngest women (15–19 years) and the oldest women (40–45 years) had significantly lower odds of contraceptive use compared to women ages 20–24 years (15–19 years AOR: 0.31; 95% CI: 0.15–0.66; 40–45 years AOR: 0.18; 95% CI 0.09–0.37).

Certainty of pregnancy and contraceptive use (Model 2):

The relationship between certainty of pregnancy and contraceptive use was significant after adjusting for sociodemographic variables: compared to reporting ‘certain’, women who reported ‘no chance/unlikely’ had 74% lower odds of using contraception (AOR: 0.26; 95% CI: 0.09–0.72). There was no significant difference in contraceptive use for women who reported ‘likely’ vs. ‘certain.’ Covariate results were similar in magnitude and significance as Model 1.

Reported infertility, certainty of pregnancy, and contraceptive use (Model 3):

With the inclusion of both main independent variables (self-reported infertility and certainty of pregnancy), each was significantly and independently associated with contraceptive use and effects were similar in magnitude and significance as the previous models. Controlling for all other covariates, women who reported infertility had 44% lower odds of using contraception than women who did not report infertility (AOR: 0.56; 95% CI: 0.39–0.83) and women who reported ‘no chance/unlikely’ that they would become pregnant after one year without using contraception had 70% lower odds of contraceptive use compared to women who were ‘certain’ they would become pregnant (AOR: 0.30; 95% CI: 0.10–0.92). All other covariates performed similarly in Model 3 as the previous models (Table 5).

Discussion

Our findings indicate that one’s fertility – both experienced infertility and perceived certainty of becoming pregnant in the future – may be consequential for contraceptive use, even when accounting for pregnancy intentions, marital status, and other sociodemographic characteristics commonly associated with contraceptive use. We found that women who experienced infertility were less likely to use contraception. The hypothesized and observed relationship between experienced infertility and contraceptive use is self-evident – if someone has experienced infertility they may, accurately or not, assume they do not need to use contraception. Behavior following this logic could be problematic, however, if one’s experience of infertility does not accurately reflect their actual chance of becoming pregnant in the future.

Following a similar logic, we found that women who said there was ‘no chance’ or it was ‘unlikely’ they would become pregnant within one year of not using contraception had significantly lower odds of current contraceptive use than women who were certain they would become pregnant. This finding aligns with constructs from existing health behavior theories – i.e., perceived susceptibility to an outcome (e.g., pregnancy) is associated with implementing a preventative behavior (e.g., contraceptive use) (Rosenstock 1974). Our study provides evidence that levels of certainty of becoming pregnant may be associated with contraceptive use, indicating that women may make contraceptive decisions based on their perception of their likelihood of becoming pregnant, which may or may not align with their biological likelihood of becoming pregnant. Indeed, among women who said that there was no chance or they were unlikely to become pregnant, under half (43%) also reported experiencing infertility. Therefore, it is possible that women are making determinations about their future fertility based on factors that may not be reliable (Gemmill, Sedlander, and Bornstein 2020).

The relationships we found between infertility, certainty of pregnancy, and contraceptive use each remained statistically significant in multivariable models. This finding supports our hypothesis that certainty of pregnancy and experienced infertility may act as independent mechanisms influencing perceived chance of becoming pregnant and contraceptive use (Figure 1).

Rates of experienced infertility in our study are similar to those found in other studies in Malawi (Rao et al. 2018; Barden-O’Fallon 2005). Given that there were no significant differences in number of children by reported infertility, it may be that experiences of infertility are episodic and do not reflect true childbearing experiences or potential. Because of the commonality of secondary infertility due to sexually transmitted infections (STIs), unsafe abortion, and previous pregnancy or delivery complications (Inhorn 2009, 2003), participants may already have children before experiencing infertility, contributing to the apparent lack of relationship between number of children and experienced infertility.

We found high rates of certainty of pregnancy in this population. One possible explanation for high rates of certainty may be that it is a “default” response that represents pronatalist norms and optimism commonly observed in Malawi (Garver 2016). The people in our sample contend with extreme poverty and uncertainty in their everyday lives, but are still likely to respond positively on measures about hopefulness and the future (Garver 2016, 2018). Given the severity of the consequences and stigma associated with not being able to become pregnant (Bornstein, Gipson, et al. 2020), it is reasonable that people erred on the side of reporting that they were likely or certain to become pregnant. It is also not surprising that certainty of pregnancy and experienced infertility were associated. Past experiences often influence future expectations.

Women who were older were more likely to report infertility. This reflects both diminished biological fertility with age, as well as a longer exposure period (exposure to pregnancy and, therefore, exposure to the possibility of not becoming pregnant) among older women. A notable bivariate finding was around divorce and infertility: 51% of women who experienced infertility had been divorced at least once, while 29% of women who had not experienced infertility had been divorced. Because the data are cross-sectional, we do not know whether divorce preceded or followed infertility; however, other findings point to abandonment and divorce as a consequence of experiencing infertility, particularly for women (Fledderjohann 2012, 2017). Along with divorce, polygamy may be considered a “solution” for a couple experiencing infertility (de Kok 2009) and, indeed, we found that 21% of women who experienced infertility were in polygamous relationships compared to 16% of those who had not (although this finding was not significant). Future interventions regarding infertility should address the potential negative social and economic consequences of divorce and complexities of polygamy (Reniers 2003; Dhont 2011), which may compound the negative consequences of infertility.

Measurement

Current contraceptive use was considerably higher in the UTHA cohort than found in other studies in Malawi. Contraceptive prevalence was 77% in the analytic sample, considerably higher than in the DHS, which reported a contraceptive use rate of 58% among married women in 2016 (NSO [Malawi] and ICF 2017). Contraceptive use may be higher in the UTHA cohort, as all participants live within the catchment area of a hospital that provides a range of free contraceptive methods. Additionally, the primary method used in our sample was injection, and we could not ascertain whether or not women received the injection every three-months. One study in Malawi found that only half of women received a second injection after first initiating the method, which could inflate contraceptive prevalence (Dasgupta, Zaba, and Crampin 2015). Furthermore, asking about contraceptive method use rather than pregnancy prevention broadly may have led to under-reporting of traditional contraceptive methods, as shown in a previous study in Ghana that found vast underreporting of traditional contraceptive use in surveys compared to in-depth interviews (Staveteig 2017). However, in the contraceptive use question, we specified that we meant modern or traditional methods. Regardless, it may be that if we probed further participants may have reported more traditional method use or if they were using their current method correctly.

While there is growing acknowledgment that perceptions of infertility may be important to how people make reproductive decisions, there are no standard or agreed upon definitions or measures of perceived infertility. Other studies have examined adjacent constructs to the certainty of pregnancy measure that we used. A recent study in Malawi looked at perceived likelihood of infertility. The authors found that just 8% of women (N=1,064) ages 21–29 years perceived they may be a little, somewhat, or very likely to be infertile, while the remaining 92% said it was not at all likely (Polis et al. 2020). This study used a measure of perceived infertility that is different from the certainty measure we used, but the findings similarly reflect high perceived fertility in Malawi. Our use of two measures – experienced infertility and perceived certainty of pregnancy – allows us to examine nuances in how people may perceive their chance of pregnancy. The first variable (infertility) captures past (or current) experiences, which are known to influence future behavior. The latter (certainty of pregnancy) captures one’s anticipated ability to become pregnant in the future, within a culturally acceptable timeframe (Bornstein, Gipson, et al. 2020).

Additional measures that studies have used related to certainty of pregnancy include percent chance of becoming pregnant (Gemmill 2018), perceived inability to become pregnant (Foster et al. 2012; Nettleman et al. 2007) and perceived speed of conception (Fledderjohann 2017), among others. While these measures are all related, it is difficult to compare across studies without a common definition and measure. Our findings suggest that certainty of pregnancy may be an important factor in contraceptive use, but in order to build the necessary body of evidence to support or refute our hypotheses around certainty and related constructs, consistent measurement is needed. After agreeing on standard definitions, it would be ideal to develop a set of psychometrically tested and standardized measures that are culturally and contextually appropriate in various settings.

Our measure of self-reported infertility in this study is limited in that someone who experienced a period of unsuccessfully trying to become pregnant for two or more years in the past – perhaps at the beginning of their reproductive life-course – is categorized the same as someone who is currently experiencing infertility. Given that infertility is not well defined in terms of exposure (i.e., timing and frequency of sex), divorce and remarriage in Malawi are common (Reniers 2003; Bertrand-Dansereau and Clark 2016), and the cause of infertility is often unknown at the couple-level (i.e., male/female/both), the possibility exists that an individual’s report of infertility may reflect the inability of their partner or past partner(s) to conceive, rather than their own inability. Thus, experiencing a period of trying to become pregnant for two or more years unsuccessfully may not be indicative of subsequent pregnancy experiences or one’s perceived certainty of pregnancy in the future.

The measure also captures both primary and secondary infertility, which are not only different etiologically, but are also likely to have differential social and relational consequences, as well as effects on contraceptive use. One can imagine that experiencing a period of infertility followed by successfully having a child or children will have a different impact on future fertility expectations and perceived need to use contraception than having never successfully become pregnant.

Our cross-sectional data, however, do not allow us to parse out causality. We cannot determine the directionality of infertility, certainty of pregnancy, and contraceptive use. While we hypothesize that infertility and certainty of pregnancy influences contraceptive use, it may also be that contraceptive use leads one to report infertility or uncertainty. Longitudinal data that reveals when women experienced a period of infertility, followed with a detailed reproductive history, will provide insight into how timing of infertility, type of infertility (i.e., primary or secondary), and certainty of pregnancy are related. Longitudinal data may also allow us to understand if women respond to the certainty question based on recent experiences or perceptions that using a contraceptive method continues to suppress fertility for a time after discontinuation.

Implications

Despite limitations, this is one of the first studies to look at the independent relationships between experienced infertility, certainty of becoming pregnant and contraceptive use. Given our findings, it appears that women may make contraceptive decisions based on their experience with infertility and how certain they are that they can become pregnant in the future. To the extent that women who do not wish to become pregnant and whose perceptions of low fertility or infertility may be incorrect they may be at risk of having an unintended pregnancy. Our findings suggest that addressing perceptions of infertility and certainty has the potential to reduce unintended pregnancy.

Programs that aim to reduce unintended pregnancy through promoting contraceptive use need to be cognizant of both experiences and perceptions of infertility and address these issues during contraceptive counseling. Women may not contracept if they believe they may not become pregnant easily or ever. Additionally, family planning programs should intentionally include women who have experienced infertility or report that they are unlikely to become pregnant. These individuals may not identify as needing contraception and self-select out of such interventions. Therefore, programs must specifically target these groups so as to not inadvertently exclude them. Including people who opt out of family planning due to experienced or perceived infertility is particularly important in the absence of access to infertility diagnostics.

Although less widely recognized, addressing infertility is an important public health goal, along with reducing unintended pregnancy (Starrs et al. 2018; Gipson, Bornstein, and Hindin 2020). Measuring the prevalence and consequences of infertility should be treated equally important as unintended pregnancy (Gipson, Bornstein, and Hindin 2020). Treatment for infertility is largely inaccessible globally, particularly in low-resource settings (Starrs et al. 2018), which magnifies the impact of infertility and, perhaps, its salience in making reproductive decisions. Moreover, focusing exclusively on unintended pregnancy undervalues the experience of infertility, as well as the role of experienced and perceived infertility in reproductive decision making.

Conclusions

Our findings indicate that experiences and perceptions surrounding fertility are associated with contraceptive behaviors, thus, potentially how women manage their fertility to reach their reproductive goals. We also found evidence that experiences of infertility and certainty of ability to become pregnant may operate independently. Our study findings underscore the need to holistically understand how experiences and perceptions of both infertility and fertility shape reproductive decision making. Such efforts are necessary to best meet the needs of women and couples in reaching their reproductive goals.

Data availability statement:

More information about the data can be found at https://u.osu.edu/utha/ Please reach out to Alison Norris (Norris.570@osu.edu) for further information regarding data availability.