Fertility Intentions and Maternal Health Behaviour During and After Pregnancy

Abstract

This study examines associations between fertility intentions and maternal health behaviours during and after pregnancy among a nationally representative sample of 3,442 women from India. Two waves of data (2005, 2012) from the India Human Development Survey were analysed to investigate the influence of unwanted births on women’s use of antenatal care, timely postnatal care, and the delivery setting using binary and ordered logistic regression, partial proportional odds models, and propensity score weighting. Fifty-eight per cent of sample births were unwanted. Regression results show that, net of maternal and household characteristics, women with unwanted births were less likely to obtain antenatal care and had fewer antenatal tests performed. Unwantedness was also associated with a lower likelihood of delivering in an institutional setting and obtaining timely postnatal care. The relationships between unwantedness and antenatal care, postnatal care, and delivery setting were robust to models accounting for propensity weighting.

Introduction

Women’s utilization of maternal health services during and after pregnancy is known to be associated with better maternal and child health outcomes, reductions in maternal and infant mortality, and improvements in women’s overall reproductive health (McDonagh 1996; Li et al. 1996; Finger 1997; WHO 2005; Mattar et al. 2007; Sines et al. 2007). Indeed, women’s health behaviours during and after pregnancy are key devices of the World Health Organization’s (WHO) Safe Motherhood Initiative (AbouZahr 2003; Freedman et al. 2007). The extant research highlights several individual- and household-level factors that influence women’s utilization of maternal health services, including women’s level of education (Celik and Hotchkiss 2000), autonomy in household decision-making (Mistry et al. 2009; Story and Burgard 2012), and the accessibility and quality of local maternal health facilities (Navaneetham and Dharmalingam 2002; Wild et al. 2010).

An important determinant of maternal and child health that has emerged in the demographic literature is the fertility intention associated with the birth (Brown and Eisenberg 1995; Gipson et al. 2008; Singh et al. 2010; Tsui et al. 2010; Sedgh et al. 2014). Nonetheless, there is limited research on the influence of fertility intentions on women’s use of health services during and after pregnancy (Joyce and Grossman 1990; Gipson et al. 2008; Tsui et al. 2010; Kost and Lindberg 2015), which is important because inadequate use of maternal health services could lead to worse health outcomes for women and their babies, including a risk for higher maternal mortality (Alkema et al. 2016). Evidence from the Global South suggests that in some settings unintended pregnancies are associated with reduced maternal healthcare utilization, such as receiving fewer antenatal check-ups than is recommended (Eggleston 2000; Marston and Cleland 2003; Singh et al. 2013; Dibaba et al. 2013) or giving birth without the assistance of a skilled birth attendant (Marston and Cleland 2003; A. Singh, Chalasani, et al. 2012). No studies to date have examined the association between the intendedness of births and women’s timely use of postnatal care, which the present study investigates.

Our study builds on the limited research examining the relationship between fertility intentions and women’s healthcare utilization during pregnancy and after birth in developing country settings, specifically India. We aim to provide several contributions. First, whereas most of the past studies of fertility intentions and women’s health behaviour in India have been concentrated in rural areas (for example P. K. Singh et al. 2012; A. Singh et al. 2013; see A. Singh, Chalasani, et al. 2012 for an important exception), our study relies on nationally representative data—the India Human Development Study (IHDS)—thus providing more generalizable results. Second, we use prospective measures of fertility intentions, focusing on births that were wanted versus those that were unwanted. Research by Koenig and colleagues (2006) from four states in rural India found that the retrospective measures of fertility intentions used by the current Demographic and Health Survey (DHS) could lead to significant underestimates of unwanted births. Thus, our prospective measure from a nationally representative sample is likely to provide more accurate estimates. Third, as we describe further below, we analyse several detailed measures of antenatal and postnatal care use that align with WHO recommendations for maternal and child health, but have yet to be explored in this context. Finally, we employ an adaptation of a propensity score weighting approach—the inverse probability weights (IPW) estimator—as a robustness check on the regression results to help determine whether differences in maternal health behaviours among women with wanted and unwanted births are due to differential maternal traits or because of differences in intention status.

India is an important setting in which to investigate these issues due to high rates of unintended fertility and poor maternal health. As reported by Singh and colleagues (2018) using data from the United Nations and the National Family and Health Survey in India, nearly 50 per cent of the estimated 48.1 million pregnancies in India in 2015 were unwanted or mistimed. According to the WHO, about 45,000 women in India died in 2015 due to preventable pregnancy-related complications, largely a consequence of the dearth of trained professionals to supervise deliveries, and inadequate antenatal services (Alkema et al. 2016).

Despite the high levels of unintended fertility in India, few studies have evaluated the influence of fertility intentions for pregnancies carried to term on subsequent maternal health behaviours. Rather, past studies in the Indian context have focused primarily on the impact of fertility intentions on the health of the resulting child. For example, studies have examined the relationship between unintendedness and child acute respiratory infection, diarrhoea (Jensen and Ahlburg 1999), stunting (A. Singh, Chalasani, et al. 2012; Upadhyay and Srivastava 2016), full vaccination by WHO standards, and child mortality (A. Singh, Chalasani, et al. 2012; Singh et al. 2013). Three studies from India have examined the influence of unintended fertility on women’s health outcomes or behaviours. First, Singh and colleagues’ (2013) study of unintended pregnancy and maternal and child health found that women reporting unwanted births were 2.32 times as likely as those reporting wanted births to obtain inadequate prenatal care. Whereas this is one of the only studies that uses prospective data to examine the linkages between pregnancy intentions, women’s use of prenatal care, and child vaccination in the context of India; it is limited to examining this relationship for residents of rural areas in four states: Bihar, Jharkhand, Maharashtra, and Tamil Nadu. Second, Singh and colleagues (A. Singh, Chalasani, et al. 2012) examined the association between unintended fertility and delivery supervision using family fixed effects to account for unobserved heterogeneity. Their findings show that mistimed births were 1.3 times as likely as wanted births to be delivered in the absence of a trained birth attendant. However, this study relied on cross-sectional data, and the measurements of birth intention were retrospective, meaning that there could be recall bias due to ex post-rationalization (Lightbourne 1985; Bongaarts 1990, 2011; Westoff 1991; Bhushan and Hill 1996). Finally, Singh and colleagues (L. Singh et al. 2012) examined the relationship between fertility intentions and antenatal care, safe delivery, and postnatal care among married adolescent (15–19) mothers in rural India. Although the authors document several correlates of maternal health behaviour, birth wantedness is not examined in the final regression models. Additionally, the measure of postnatal care only assesses whether women obtained one visit within 42 days of the birth. Our study extends this past work by using longitudinal data with prospective measurement of fertility intentions from India’s first nationally representative panel survey (across both rural and urban areas). In addition, our study is the first in the Indian context to examine the relationship between women’s fertility intentions and a detailed measure of postnatal care use that captures whether the visit meets the WHO recommendations for timely care (within 24 hours of the birth).

Context

India is the second most populous country in the world and is likely to surpass China by 2027 to become the most populous country in the World (United Nations Population Division 2019). Despite its growing population, India has seen a marked decline in fertility over the past several decades: the total fertility rate (TFR) dropped from 5.9 children per woman in 1960 (World Bank 2017) to 2.2 in 2016 (International Institute for Population Sciences (IIPS) and ICF 2017), though there is considerable variation in fertility across different states and regions. According to the results of the India National Family Health Survey (IIPS and ICF 2017), the total wanted fertility rate in India is 1.8 children per women. Nearly 50 per cent of married women in India use modern contraceptives, and 13 per cent have an unmet need for contraception. Female sterilization is the most popular method of contraception in India, and is used by around 36 per cent of married women aged 15–49 (IIPS and ICF 2017).

Methods

Data

The data come from two rounds of the India Human Development Survey (2005 and 2012). The 2005 survey includes data collected in face-to-face interviews with individuals in 41,554 households across 33 (now 34) States and Union territories, covering 1,503 villages and 971 urban regions in India (Desai et al. 2010). Follow-up interviews were conducted in 2012 with 83 per cent of the IHDS I households and split households (separated from the root household) that resided in the same community. Attrition was lower among larger, rural households, particularly those that owned land (Thorat et al. 2017). Questions in the household module (e.g., questions on income, consumption, social capital) were answered by heads of household—often men—whereas the health (e.g., questions on fertility history, ideal number of children) and education modules were answered by one ever-married woman per household, often the spouse of the household head. In the present study our sample is limited to non-pregnant, currently married women, aged 18–40 in 2005, who participated in both surveys and had at least one birth between 2005 and 2012.

Sample

The sample for the current study was drawn from data for 25,479 women who participated in both surveys. Of these women, 18,737 met the sampling criteria of being non-pregnant, currently married, and aged 18–40 in 2005. Of these women, 9.5 per cent (n=1,783) had invalid or missing data on fertility intentions and were dropped from the sample. Because our analysis focuses on whether a birth was wanted or unwanted, women had to have a birth in the interim survey period; therefore, we restricted the sample to women who had at least one birth between 2005 and 2012 (as reported in 2012). We used two measures to determine whether women had a birth in the interim survey period: first, we subtracted the number of children ever born as reported in 2005 from the number of children ever born as reported in 2012. Second, we examined the number of children born after January 2005 as reported in 2012. We dropped <1 per cent (n=17) of women because of missing data on the first measure of the number of children we calculated. Of the remaining 16,937 women eligible for the sample, 13,264 (78 per cent) did not have a birth in the interim period¹ and were thus dropped, reducing the sample to 3,673 women. We took a conservative approach and further restricted our analysis to cases where there was consistency in the number of children reported across both measures, and thus whether the most recent birth would be labelled as wanted or unwanted (211 women were dropped for inconsistent responses). The analytic sample was also limited to women with non-missing data on all independent variables. Therefore, we dropped <1 per cent (n=20) of the sample due to missing data on education, which was the only independent variable with missing values. After dropping these cases, the final study sample consists of 3,442 women. The sample sizes for each model range from 3,153 to 3,345 due to missing data on the dependent variables (see Table 1).

Table 1.

Weighted summary statistics

	N	%/Mean (sd)
Dependent variables
Any antenatal check-up	3,345	81.9
Adequate antenatal check-ups	3,299	39.3
Antenatal check-up index	3,153
No tests done (0)		17.1
All tests done (8)		12.1
Postnatal care index	3,320
No check-up		38.5
First check-up after 24 hours of birth		24.7
First check-up within 24 hours of birth		36.8
Delivery in institutional setting	3,326	59.5
Independent variables
Unwanted birth	3,442	58.1
Age	3,442	24.58 (4.83)
Number of living children	3,442	1.77 (1.26)
Household asset quintile
Poorest	3,442	22.9
Second quintile	3,442	21.7
Third quintile	3,442	24.7
Fourth quintile	3,442	16.9
Richest	3,442	13.8
Caste group
Scheduled Castes (SC)	3,442	24.7
Scheduled Tribes (ST)	3,442	7.5
Other Backward Classes (OBC)	3,442	44.4
Forward Castes (FC)	3,442	23.4
Religion
Hindu	3,442	79.9
Muslim	3,442	15.5
Other religion	3,442	4.7
Urban	3,442	19.9
Education
Illiterate	3,442	47.8
Incomplete primary	3,442	6.5
Primary	3,442	28.7
Secondary	3,442	8.4
Higher secondary	3,442	5.1
College and higher	3,442	3.4
Empowered Action Group (EAG) state	3,442	58.7

Measures

Dependent variables.

We analyse various aspects of women’s use of antenatal care, characteristics of their baby’s delivery, and women’s use of postnatal care to investigate women’s health behaviour during and after their pregnancy. In 2012, all women who had at least one birth between 2005 and 2012 were asked about their use of antenatal and postnatal care and the setting of their delivery for their most recent birth since January 2005. Between 2005 and 2012, 55 per cent of the 3,442 women in the sample had one child, 31 per cent had two children, and 14 per cent had three or more children. Our analysis focuses on the most recent birth².

We analyse five dependent variables related to maternal health behaviour during and after pregnancy. a) Any antenatal check-up: This variable measures whether women obtained any antenatal check-ups during their pregnancy. The variable takes a value of 0 if the woman obtained no antenatal care, and 1 if she obtained at least one antenatal check-up during her pregnancy. b) Adequate antenatal check-ups: The WHO recommends at least 4 antenatal check-ups during pregnancy (WHO 2006). To assess whether women obtained the WHO-recommended number of antenatal check-ups during their pregnancy, the variable takes a value of 1 if the woman had four or more antenatal check-ups during pregnancy and 0 otherwise. c) Antenatal check-up index: This is an additive index counting the number of different tests that the woman received during her pregnancy. Referring to their most recent birth since January 2005, women were asked: ‘Did you have the following performed at least once during any of your antenatal check-ups for this pregnancy? i) weight check-up, ii) blood test, iii) sonogram, iv) urine test, v) BP check, vi) amniocentesis, vii) internal check-up, viii) abdominal examination.’ The index values range from 0 to 8, with 0 indicating that no tests were performed and 8 indicating that all of the tests were performed at least once during the pregnancy. d) Delivering in an institutional setting: The WHO recommends delivery under the assistance of a trained person (such as a doctor, nurse, or health professional) or in an institutional setting (such as a hospital, private clinic, or nursing home) as standards for safe delivery (Department of Making Pregnancy Safer 2007). This variable takes a value of 1 if the delivery took place in a government hospital/clinic, private nursing home, or some other institutional setting with health personnel; and it takes a value 0 if the delivery took place at home. e) Postnatal care index: This variable is an indicator of women’s postpartum healthcare utilization, which is important for both maternal and infant health. WHO guidelines for postnatal care recommend that the first postnatal check-up occur within 24 hours of the birth, regardless of the place of birth (WHO 2014). The postnatal care index takes a value of: 0 if the woman or her child had no postnatal check-up; 1 if the woman or her child had a postnatal check-up more than 24 hours after the birth but within 2 months; and 2 if the women or her child had a postnatal check-up within 24 hours of the birth.

Independent variables.

The key independent variable measures women’s prospective fertility intentions. We assess whether a woman’s most recent birth was wanted versus unwanted by comparing the number of additional desired children in 2005 with the number of children born between 2005 and 2012. If the number of additional desired children in 2005 was less than the number of children born between 2005 and 2012 (including those who died in the interim), then the most recent birth was labelled as unwanted. If the number of additional desired children in 2005 was greater than or equal to the number of children born between 2005 and 2012, the most recent birth was labelled as wanted (also see Yeatman and Sennott 2015). This variable takes a value of 1 if the last birth was unwanted, and 0 if the last birth was wanted. Due to a lack of data on fertility timing preferences, our analysis focuses on differences between wanted and unwanted births and does not account for mistimed births.

We also assess several variables measuring individual- and household-level characteristics in 2005 that have been shown to be important in past research on maternal health behaviours. First, socio-demographic traits are important determinants of maternal health behaviour during and after pregnancy. For example, studies from developing countries have found women who are older, have higher parity, are less educated, and belong to poorer households have lower likelihoods of adequate maternal healthcare utilization (Navaneetham and Dharmalingam 2002; Sharma 2004; Chandhiok et al. 2006; Simkhada et al. 2008; Ahmed et al. 2010; Amin et al. 2010; Pathak, et al. 2010; A. Singh, Chalasani, et al. 2012; A. Singh, Padmadas, et al. 2012). In the present study we control for the woman’s age (18–40), which is coded continuously because models examining an age-squared term indicated it was not significant. We also control for the number of living children a woman had in 2005, her education level (some primary education (1–4), primary complete (5–9), secondary complete (10–11), higher secondary complete (12 and some college), and college degree and more), and her household asset quintile (five dummy variables for each quintile ranging from the poorest to the richest). In India specifically, maternal healthcare utilization varies by several other factors. First, women belonging to lower castes such as Scheduled Castes (SC), Scheduled Tribes (ST), and Other Backward Classes (OBC) are less likely to utilize adequate maternal healthcare facilities such as safe delivery, and adequate postnatal and antenatal care, compared to those belonging to higher castes (Navaneetham and Dharmalingam 2002; Pallikadavath ey al. 2004; Matthews et al. 2005; A. Singh, Chalasani, et al. 2012; P. K. Singh et al. 2012). In the present study we control for the woman’s caste group by including four dichotomous variables indicating membership in a Forward Caste group (FC), SC, ST, or OBC. Second, some studies have found that Muslims are less likely to utilize safe delivery care compared to Hindus in India (Navaneetham and Dharmalingam 2002; P. K. Singh et al. 2012), whereas other studies have found mixed results for religious groups (Sugathan et al. 2001). We control for women’s religious group using three dichotomous variables: Hindu, Muslim, or other religion. Third, a woman’s region of residence is important to consider in order to better understand socio-demographic processes in India. Specifically, a group of Northern States—including Bihar, Chhattisgarh, Jharkhand, Madhya Pradesh, Orissa, Rajasthan, Uttarakhand, and Uttar Pradesh—have higher populations, lower educational attainment, lower status of women, less adequate administration, and a larger prevalence of traditional norms and beliefs. These States have been termed as “Empowered Action Group (EAG)” States and are the focus of various government health and family planning programs. These States accounted for 46 per cent of India’s population in 2011 and 53 per cent of the growth in population (Registrar General India 2011). Past studies have found that women residing in Northern EAG States were less likely to use maternal healthcare services compared to women in other areas (International Institute for Population Sciences Macro International 2007; P. K. Singh et al. 2012; L. Singh et al. 2012). Therefore, in the present study we include a binary variable for state of residence that takes a value of 1 if a woman lives in an EAG State and 0 if she lives in a non-EAG State. In addition, we control for whether a woman lives in an urban (1) versus a rural area (0).

Analyses

We first calculate descriptive statistics for the variables, including percentages, means, and standard deviations. Next, we analyse two models for each dependent variable. The first model examines the bivariate relationship between the wantedness of a birth and maternal healthcare utilization as measured by women’s antenatal and postnatal care utilization and whether the delivery occurred in an institutional setting. Maternal healthcare utilization varies by both socioeconomic and demographic characteristics, however, the characteristics of the woman and her household could also impact the wantedness of the birth and thus serve as confounding factors. Previous research has shown that fertility intentions and subsequent maternal health behaviour are associated with various social, demographic, and economic variables such as: age, education, household assets, parity, and area of residence (rural vs. urban) (Dommaraju and Agadjanian 2009; Morgan and Rackin 2010; Kodzi et al. 2010; Hayford and Agadjanian 2012; Dharmalingam, et al. 2014). To account for this, in a second set of models, we assess whether the socio-demographic traits of the woman and her household mediate the relationship between the wantedness of the birth and a woman’s subsequent health behaviour using stepwise weighted logistic regression for binary outcome variables (i.e., any antenatal care, adequate antenatal care, and delivery in an institutional setting). The antenatal check-up index and postnatal care index are ordered outcome variables. While ordered logit regression is an option for these variables, it is important to test whether a critical assumption of the ordered logit model is violated in the data (the parallel lines assumption). Results from the Brant test indicate that for the antenatal check-up index, the parallel lines assumption is not violated and an ordered logit model can be used. However, for the postnatal care index, the parallel lines assumption is violated. Thus, for the postnatal care index we use the partial proportional odds model, which relaxes the parallel lines assumption for some independent variables (Williams 2016; May and Reynolds 2018). In order to determine which variables should be left unconstrained in the partial proportional odds model we used STATA’s gologit2 command with the autofit option and an α of 0.01 (Williams 2006). Next we used the fitstat option in STATA to compute the BIC for the partial proportional odds model and the generalised ordered logit model. We found that the partial proportional odds model (with sample weights) was the best-suited model for the postnatal care index outcome variable.

Recent studies on fertility intentions and maternal health from the United States and India have used fixed effects models (Joyce et al. 2000; Barber and East 2009; Guzzo and Hayford 2012; A. Singh, Chalasani, et al. 2012) and propensity score matching (Kost and Lindberg 2015) to correct for selection bias. Propensity score analyses are less sensitive to model specification errors compared to regression models (Drake 1993; Dehejia and Wahba 2002; Messer et al. 2010; Stuart 2010; McCaffery et al. 2013; Kost and Lindberg 2015). Therefore, in the present study we use an adaptation of propensity score matching (PSM)—the IPW estimator—as a robustness check to the stepwise regression models. This adjustment is useful for disentangling the impact of a woman’s fertility intentions from the impact of her other characteristics on her health behaviours. The IPW method estimates the parameters of the treatment model (the model predicting the wantedness of a birth), and calculates estimated inverse probability weights. Thereafter, the estimated inverse probability weights are used to calculate the predicted probabilities and predicted counts for the outcomes measuring maternal health behaviours for mothers who have unwanted and wanted births (see Cattaneo, 2010). Estimated predicted probabilities and predicted counts from IPW analyses help model a counterfactual condition that shows what the variation between mothers with wanted and unwanted births would be if they had the same likelihood of belonging to the groups in which we find them (Kost and Lindberg 2015). In addition, we included sample weights: each observation’s inverse probability weight (IPW) is multiplied by the sample weights in order to obtain unbiased effects based on the population of all births in India (see DuGoff et al. 2013; Kost and Lindberg 2015).

The propensity scores used for weighting are estimated from a binomial logistic regression model with women’s fertility intentions (unwanted birth = 1) serving as the dependent variable. The independent variables included maternal and household characteristics that could be related to fertility intentions: age, education level, number of living children in 2005, caste, religion, household asset quintile, region (EAG vs. non-EAG State), and area of residence (urban vs. rural). Next, we estimate three sets of binomial logistic regression models and two sets of Poisson regression models analysing the relationship between fertility intentions and the binary (any antenatal care, adequate antenatal care, and delivery in an institutional setting), and ordered (antenatal check-up index and postnatal care index) dependent variables measuring maternal health behaviour. For the antenatal and postnatal care indices we use Poisson regression because it is the only suitable option for ordered outcome variables using IPW in Stata. This enables us to compare the predicted probabilities and predicted counts from the unadjusted data (using only sample weights) to those of the adjusted sample (weighing each observation by the inverse of its probability and sample weights). All data were analysed using Stata 15 (Stata Corp 2017).

Results

Table 1 shows the descriptive statistics for the sample. Unwanted births comprised the majority of the births in the sample: around 58 per cent of women’s most recent births were classified as unwanted and 42 per cent as wanted. To further explore the circumstances under which women had unwanted births, around 42 per cent women in the analytic sample who had an unwanted birth did not want any more children in 2005, and 39 per cent of the women who wanted only one (more) child in 2005 exceeded their desired fertility by having at least two by 2012 (not shown).

Around 82 per cent of women in the sample obtained at least one antenatal check-up during pregnancy, however only 39 per cent obtained the recommended number of check-ups (four or more) as determined by the WHO. The mean score on the antenatal check-up index was 4.3; thus, on average women received four of the eight tests during their pregnancy. Around 60 per cent of women in the sample delivered in an institutional setting whereas around 40 per cent delivered at home. Almost 39 per cent of women in the sample obtained no postnatal check-ups (for themselves or their baby); 25 per cent obtained at least one postnatal check-up more than 24 hours after birth but within 2 months; and 37 per cent of women in the sample received their first postnatal check-up within 24 hours of the birth, therefore meeting the WHO recommendations for timely postnatal care. The average age of women in the sample was around 25 and women typically had between one and two children in 2005. Nearly one in four (23 per cent) women in the sample belonged to the poorest households and 14 per cent of women belonged to the richest households. Twenty-three per cent of women belonged to Forward Caste groups and the rest belonged to SC, ST, and OBC groups. The majority of the women in the sample were Hindus (80 per cent), whereas nearly 16 per cent were Muslims, and 5 per cent belonged to other religions. One in five women (20 per cent) in the sample resided in urban areas. Educational attainment was very low: almost half (48 per cent) of women in the sample were illiterate and only 3 per cent had obtained a college degree. Finally, 59 per cent of women resided in the EAG States, which are less developed and growing faster compared to the non-EAG States.

Table 2 shows maternal health behaviours by the wantedness of a woman’s most recent birth in 2012. There are strong, consistent differences for women whose most recent birth was unwanted versus wanted. Seventy-five per cent of women who had an unwanted birth had at least one antenatal check-up compared to 92 per cent of women who had a wanted birth. Around 32 per cent of those who had an unwanted birth obtained the adequate number of antenatal check-ups as recommended by the WHO compared to about half of the women whose most recent birth was wanted. For women who had unwanted births, 24 per cent had zero tests done and 6 per cent obtained all eight tests during their pregnancy. In contrast, nearly 9 per cent of those who had a wanted birth had zero tests done whereas almost one in five (20 per cent) obtained all eight tests during their pregnancy. Forty-five per cent of women who had an unwanted birth obtained no postnatal check-ups; whereas amongst women who had wanted births only 30 per cent of women obtained no postnatal check-ups. In contrast, 44 per cent of women who had a wanted birth obtained the first postnatal check-up within 24 hours of the birth compared to only 32 per cent of women who had an unwanted birth. Finally, half of those who had an unwanted birth delivered in an institutional setting, compared to 72 per cent of women who had a wanted birth. These results highlight the strong binary relationships between women’s fertility intentions and their health behaviours during and after pregnancy.

Table 2.

Maternal health behaviours by fertility intentions in India (weighted %)

	Wanted birth	Unwanted birth
Any antenatal check-up	91.5	75.0
Adequate antenatal check-ups	49.7	31.9
Antenatal check-up index
No tests done	8.5	23.7
All tests done	19.8	6.3
Postnatal care index
No check-up	29.5	45.1
First check-up after 24 hours	26.8	23.1
First check-up within 24 hours	43.7	31.8
Delivery in institutional setting	71.9	50.6

Table 3 shows results from stepwise weighted logistic and ordered logistic regression models, examining the relationship between birth wantedness and three binary indicators of maternal health behaviour (any antenatal care, adequate antenatal care, and delivery in an institutional setting), and one of the ordered indicators of maternal healthcare utilization (antenatal check-up index), respectively, accounting for characteristics of the woman and her household. Model 1 for each of the dependent variables shows that women who had an unwanted birth were significantly less likely to avail the recommended antenatal and delivery care compared to those whose birth was wanted. After accounting for the control variables in Model 2, the values of the coefficients for the relationships between wantedness and maternal health behaviours in each set of models were attenuated, suggesting that maternal and household characteristics explain some of these relationships. However, even after controlling for all maternal and household characteristics, having an unwanted birth remains a significant predictor of obtaining any antenatal care (p<0.05), obtaining the recommended number of antenatal tests (p<0.01), and delivering in an institutional setting (p<0.01).

Table 3.

Logistic and ordered logistic regression models examining fertility intentions and maternal health behaviours in India

	Any antenatal check-up		Adequate antenatal check-ups		Antenatal check-up index		Delivery in institutional setting
Variables	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2
Unwanted birth	−1.274***	−0.458*	−0.744***	−0.122	−0.946***	−0.298**	−0.917***	−0.471**
	(0.1699)	(0.1992)	(0.1213)	(0.1501)	(0.1073)	(0.1154)	(0.1246)	(0.1535)
Age		−0.005		0.011		0.021		0.018
		(0.0201)		(0.0151)		(0.0143)		(0.0161)
Number of living children		−0.328***		−0.204**		−0.275***		−0.204**
		(0.0865)		(0.0730)		(0.0505)		(0.0676)
Household asset quintile
Second quintile		0.288		0.432*		0.212		0.183
		(0.2006)		(0.1913)		(0.1616)		(0.1765)
Third quintile		0.810***		0.958***		0.600***		0.288
		(0.2445)		(0.2143)		(0.1613)		(0.2038)
Fourth quintile		1.410***		0.602**		0.761***		0.399+
		(0.3006)		(0.2153)		(0.1752)		(0.2176)
Richest		1.033*		0.502*		0.838***		0.919**
		(0.4078)		(0.2502)		(0.2023)		(0.2957)
Caste Group
Scheduled Castes (SC)		0.263		−0.128		−0.056		−0.110
		(0.2396)		(0.1700)		(0.1477)		(0.1843)
Scheduled Tribes (ST)		1.001***		0.073		0.209		−0.719**
		(0.2902)		(0.2717)		(0.2008)		(0.2605)
Other Backward Classes (OBC)		−0.048		0.209		0.209+		−0.187
		(0.2103)		(0.1506)		(0.1246)		(0.1584)
Religion
Muslim		−0.041		−0.154		−0.179		−0.635**
		(0.2446)		(0.2003)		(0.1502)		(0.1942)
Other Religion		−0.518		0.241		−0.507**		−0.659**
		(0.3321)		(0.2509)		(0.1753)		(0.2470)
Urban		0.303		0.331*		0.467***		0.882***
		(0.1951)		(0.1337)		(0.1096)		(0.1438)
Education
Incomplete Primary		0.298		0.145		0.273		−0.176
		(0.3347)		(0.2277)		(0.1806)		(0.2160)
Primary		0.891***		0.712***		0.870***		0.318+
		(0.2155)		(0.1708)		(0.1377)		(0.1650)
Secondary		0.818+		0.824**		0.811***		0.651*
		(0.4233)		(0.2561)		(0.1865)		(0.2880)
Higher Secondary		1.368*		1.151***		0.885***		0.809**
		(0.6312)		(0.2752)		(0.2188)		(0.3080)
College and higher		2.774**		1.762***		1.285***		3.341**
		(0.8555)		(0.3196)		(0.2273)		(1.0556)
Empowered Action Group (EAG) State		−1.280***		−1.218***		−1.817***		−0.636***
		(0.1989)		(0.1279)		(0.1154)		(0.1353)
Ancilliary Parameters++		0020
cut1					−2.179***	−2.371***
					(0.0989)	(0.3745)
cut2					−1.611***	−1.667***
					(0.1021)	(0.3671)
cut3					−1.339***	−1.303***
					(0.1027)	(0.3627)
cut4					−1.118***	−0.993**
					(0.0985)	(0.3620)
cut5					−0.777***	−0.509
					(0.0910)	(0.3612)
cut6					−0.262**	0.227
					(0.0869)	(0.3592)
cut7					0.493***	1.284***
					(0.0875)	(0.3770)
cut8					1.519***	2.598***
					(0.1034)	(0.3705)
Constant	2.374***	2.630***	−0.014	−0.696+			0.942***	0.694+
	(0.1506)	(0.5446)	(0.0875)	(0.3979)			(0.1007)	(0.4190)
Sample Size	3, 345	3,345	3,299	3,299	3,153	3,153	3,326	3,326

Robust standard errors in parentheses

^***p<0.001

^**p<0.01

^*p<0.05,

⁺p<0.10

^++:These are cut points that are used to differentiate the adjacent categories of the antenatal check-up index. For example, cut1 is the estimated cut point on the latent variable that differentiates those with no tests performed during antenatal check-ups from those with a score of 1, 2, 3, 4, 5, 6, 7 or 8 on the antenatal check-up index, when values of the independent variables are zero.

As shown in Table 3, several of the control variables were significantly associated with women’s maternal health behaviours in each model. The number of living children was negatively associated with each dependent variable. Household assets were generally positively related to each outcome variable. Being from a Scheduled Tribe was positively associated with receiving any antenatal care and negatively associated with delivering in an institutional setting. Compared to Hindu women, Muslim women and those belonging to other religions were less likely to deliver in an institutional setting. Women of other religions were also received lower scores on the antenatal check-up index. Living in an urban area (compared to a rural area) was positively associated with receiving an adequate number of antenatal check-ups, obtaining higher scores on the antenatal check-up index, and delivering in an institutional setting. Education was generally positively associated with women’s use of antenatal care and delivering in an institutional setting, and being from an EAG state was negatively associated with all outcome variables.

Table 4 shows the results from stepwise partial proportional odds models (with sample weights) examining the relationship between birth wantedness and the postnatal care index. The first panel in Models 1 and 2 (>0) compares women who have a score of 1 and 2 on the postnatal care index compared to those who have a score of 0 on the index, whereas the second panel (>1) compares women who have a score of 2 with those who have a score of 0 and 1 on the index. Results show that women who had unwanted births had lower scores on the postnatal care index compared to those who had wanted births. In other words, women who had unwanted births were less likely to obtain timely postnatal care compared to women whose births were wanted. This relationship remains significant after controlling for maternal and household characteristics in Model 2 (p<0.001).

Table 4.

Partial proportional odds regression models examining fertility intentions and timely postnatal care in India

	Postnatal care index		Postnatal care index
	Model 1		Model 2
Variables	>0	>1	>0	>1
Unwanted birth	−0.588***	−0.588***	−0.438***	−0.438***
	(0.1049)	(0.1049)	(0.1144)	(0.1144)
Age			−0.007	−0.007
			(0.0123)	(0.0123)
Number of living children			−0.047	−0.047
			(0.0515)	(0.0515)
Household asset quintile
Second quintile			0.032	0.032
			(0.1614)	(0.1614)
Third quintile			0.548**	0.548**
			(0.1702)	(0.1702)
Fourth quintile			0.205	0.205
			(0.1766)	(0.1766)
Richest			0.459*	0.459*
			(0.2030)	(0.2030)
Caste group
Scheduled Castes (SC)			−0.014	−0.014
			(0.1514)	(0.1514)
Scheduled Tribes (ST)			−0.460*	−0.460*
			(0.2070)	(0.2070)
Other Backward Classes (OBC)			0.119	0.119
			(0.1242)	(0.1242)
Religion
Muslim			−0.278+	−0.278+
			(0.1452)	(0.1452)
Other religion			0.085	0.085
			(0.1863)	(0.1863)
Urban			0.352***	0.352***
			(0.1046)	(0.1046)
Education
Incomplete primary			−0.232	−0.232
			(0.2035)	(0.2035)
Primary			0.183	0.183
			(0.1332)	(0.1332)
Secondary			0.337+	0.337+
			(0.1977)	(0.1977)
Higher secondary			−0.066	−0.066
			(0.2096)	(0.2096)
College and higher			0.389+	0.389+
			(0.2305)	(0.2305)
Empowered Action Group (EAG) state			−0.172	0.622***
			−0.1218	−0.1283
Constant	0.815***	−0.214*	0.756*	−0.793*
	(0.0829)	(0.0845)	(0.3524)	(0.354)
N	3,320	3,320	3,320	3,320

Robust standard errors in parentheses

^***p<0.001

^**p<0.01

^*p<0.05

⁺p<0.10

Note: The first panel in Models 1 and 2 (>0) compares women who have a score of 1 or 2 on the postnatal care index to those who have a score of 0; the second panel (>1) compares women who have a score of 2 on the index with those who have a score of 0 or 1.

Model 2 shows that several control variables were significantly associated with the postnatal care index. Household assets generally had a positive association with the use of postnatal care whereas being from a Scheduled Tribe had a negative association. Muslim women were less likely to obtain timely postnatal care compared to Hindu women (marginally significant, p<0.10). Living in an urban area was positively associated with women’s use of timely postnatal care, as was attending secondary school (marginally significant) and completing college (marginally significant) Finally, living in an EAG state was positively associated with having a postnatal care check-up within 24 hours of the birth³.

Table 5 shows the predicted probabilities for Models 1 and 2 for each of the dependent variables in Tables 3 and 4 by the wantedness of the birth. The probability of obtaining at least one antenatal check-up for women who had an unwanted birth between 2005 and 2012 was 0.75 (panel 1, model 1); it increased to 0.80 (panel 1, model 2) after taking the control variables into account. In contrast, the probability of a woman obtaining at least one antenatal check-up when she had a wanted birth was 0.92 before taking into account other traits of the woman and her household; it decreased to 0.86 after controlling for other factors. Though the difference between obtaining any antenatal check-ups for those who had an unwanted birth versus a wanted birth decreased after taking into account maternal and household characteristics, women who had an unwanted birth still had significantly lower probabilities of obtaining at least one antenatal check-up (p<0.05). The probability of a woman who had an unwanted birth obtaining four or more antenatal check-ups was 0.32 (panel 2, model 1), whereas for those who had a wanted birth it was 0.50, a significant difference (p<=0.001). However, after taking into account maternal and household characteristics, there was not a significant difference in the probability of obtaining the WHO-recommended number of antenatal check-ups by women’s fertility intentions (panel 2, model 2). The probability of a mother who had an unwanted birth having all eight tests performed during antenatal check-ups was 0.08 (panel 3, model 1), whereas the probability for those who had a wanted birth obtaining all eight tests was 0.18. This difference decreased after taking the control variables into account because women who had lower SES, less education, more living children, and resided in rural area and EAG states had lower probabilities of obtaining tests during antenatal check-ups. Nonetheless, women who had unwanted births still had significantly lower probabilities of obtaining tests during antenatal check-ups compared to those who had wanted births (p<0.01) (panel 3, model 2). The probability of a mother who had an unwanted birth obtaining a postnatal check-up within 24 hours of the birth was 0.31 (panel 4, model 1), whereas the probability for those who had a wanted birth was 0.45. After taking into account maternal and household characteristics this difference decreased, however, women who had an unwanted birth still had a lower probability of receiving timely postnatal care compared to those who had a wanted birth (p<0.001) (panel 4, model 2). Finally, the probability of delivering in an institutional setting for women who had an unwanted birth was 0.51 compared to 0.72 for women who had a wanted birth (panel 5, model 1). Although the difference between women with an unwanted and a wanted birth decreased in Model 2, the probability of delivering in an institutional setting was still significantly lower for those who had an unwanted birth compared to those who had a wanted birth (p<0.01) (panel 5, model 2).

Table 5.

Predicted probabilities of maternal health behaviours by fertility intentions in India

	Any antenatal check-up		Adequate antenatal check-ups		Antenatal check-up index (full score)		Postnatal care index (first check-up within 24 hours of birth)		Delivery in institutional setting
	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2
Wanted birth	0.915	0.856	0.497	0.406	0.18	0.134	0.447	0.423	0.719	0.653
Unwanted birth	0.75	0.804	0.319	0.383	0.078	0.108	0.31	0.326	0.506	0.56

Table 6 shows the predicted probabilities (for the binary indicators of maternal healthcare utilization) and predicted counts (for the antenatal and postnatal care indices) that we would find if women who had wanted and unwanted births had similar distributions of socio-demographic characteristics. The IPW adjustment helps disentangle the impact of a woman’s fertility intentions from the impact of her other traits on her maternal health behaviours. The results are largely consistent after using the IPW estimator. Women who had unwanted births remained significantly less likely to obtain any antenatal check-up, deliver in an institutional setting, and to have a lower score on the indices for antenatal and postnatal care compared to women whose births were wanted. Moreover, the difference between women who had wanted and unwanted births on the postnatal care index remained significant even if antenatal care was included as an explanatory variable in the model (not shown).

Table 6.

Predicted probabilities and predicted counts for women with unwanted and wanted births with and without using IPW estimator

	Wanted birth	Unwanted birth
Without IPW
Any antenatal check-up	0.856*	0.804*
Adequate antenatal check-ups	0.406	0.383
Antenatal check-up index	4.369**	4.151**
Postnatal check-up index	1.09***	0.899***
Delivery in institutional setting	0.653**	0.56**
Adjusted with IPW
Any antenatal check-up	0.851*	0.802*
Adequate antenatal check-ups	0.424	0.388
Antenatal check-up index	4.435+	4.17+
Postnatal check-up index	1.125***	0.924***
Delivery in institutional setting	0.65**	0.557**

Note: Models using IPW do not allow for ordered outcome variables. Therefore, in this table we treat both the antenatal and postnatal care indices as count variables and use Poisson regression in order to compare results with and without the IPW adjustment. The only suitable option for ordered outcome variables using IPW in Stata is Poisson Regression.

Discussion

Our results show that women’s fertility intentions have a significant influence on maternal healthcare utilization in India, even after taking into account maternal and household characteristics that are associated with both fertility intentions and women’s healthcare utilization. Specifically, we find that women who have unwanted births are significantly less likely than women whose births were wanted to obtain any antenatal check-ups, to receive the recommended tests during pregnancy, to deliver in an institutional setting, and to obtain timely postnatal care. However, since individual and household characteristics could be linked to both the wantedness of a birth and maternal health behaviours, it can be difficult to determine if a regression model is specified correctly (Kost and Lindberg 2015). This is important because if the model is misspecified, inferences about the relationships between birth wantedness and maternal health behaviours could be inaccurate, particularly due to differences in maternal and household characteristics for mothers with wanted versus unwanted births. Evaluating the relationships between the wantedness of a birth and maternal health behaviours after using propensity score weighting—an approach that renders mothers with wanted and unwanted births more similar in terms of observed maternal and household characteristics—reduces the sensitivity to model specification errors compared to regression models for maternal health behaviours (Drake 1993; Dehejia and Wahba 2002; Messer et al. 2010; Stuart 2010; McCaffrey et al. 2013; Kost and Lindberg 2015). After using the IPW estimator, we find that women who have an unwanted birth are significantly less likely to obtain any antenatal check-ups, have fewer tests performed during antenatal check-ups, deliver less often in an institutional setting, and are less likely to obtain timely postnatal care, effects that are robust to maternal and household characteristics.

The extant research from the Global South examining the association between fertility intentions and women’s health behaviours has shown mixed results. Some studies have found that unintendedness is associated with beginning antenatal care later and failing to obtain the recommended number of antenatal care visits (Eggleston 2000; Barrick and Koenig 2008; Bassani et al. 2009). Other studies have found an inconsistent relationship between fertility intentions and antenatal care utilization (Gage 1998; Marston and Cleland 2003). Thus, scholars have called for additional research on the impact of unwanted births on various dimensions of maternal healthcare in developing country settings (Gipson et al. 2008). Our study responds to this call, building on this body of past research in several ways. First, we rely on a prospective measure of fertility intentions that is able to capture a woman’s desire for future fertility prior to when a pregnancy occurs and therefore avoids the potential bias in retrospective measures (Koenig et al. 2006; Yeatman and Sennott 2015). Second, we use two waves of data from the first nationally representative study from India, which is both generalizable and ideal for investigating the influence of fertility intentions on women’s subsequent health behaviours. Finally, our study is the first to examine the association between fertility intentions and women’s use of timely postnatal care—as recommended by the WHO—while also accounting for possible selection bias. Together, our results highlight an important factor—maternal healthcare utilization—that could negatively influence both a woman and her baby’s health during an unwanted pregnancy and even after birth, as under-utilization of healthcare services is associated with poor health outcomes (Dibaba et al. 2013).

Our study found that several maternal and household socio-demographic characteristics were associated with women’s healthcare utilization. First, compared to the poorest women, those belonging to the richest households (and households in the fourth asset quintile) were more likely to avail adequate antenatal and postnatal care and to deliver in an institutional setting. These results suggest that these women may find health facilities more affordable and accessible, consistent with other research from a variety of developing countries (Miles-Doan and Brewster 1998; Ahmed et al. 2010), and specifically in the Indian context (Pathak et al. 2010; Kesterton et al. 2010; A. Singh, Chalasani, et al. 2012; A. Singh, Padmadas, et al. 2012; P. K. Singh et al. 2012; L. Singh et al. 2012). Second, compared to women who were illiterate, women who had more education (those who had secondary, higher-secondary, or college degrees) were more likely to use antenatal care facilities, deliver in institutional settings, and obtain timely postnatal care (Sunil et al. 2006; Amin, Shah, and Becker 2010; Ahmed et al. 2010; Kesterton et al. 2010; A. Singh, Chalasani, et al. 2012; P. K. Singh et al. 2012). This may be because educated women are more likely to communicate with their husbands and other family members on issues linked to health (Navaneetham and Dharmalingam 2002; P. K. Singh et al. 2012), which could lead to higher levels of support for seeking health services when needed. Further, women with higher education are also more likely to have access to higher quality healthcare facilities and, in general, to more often utilize healthcare facilities because they are more aware of the benefits (Celik and Hotchkiss 2000; P. K. Singh et al. 2012). Third, we found that women residing in urban areas were more likely than those in rural areas to obtain antenatal care. This is in contrast to past research from India that found no differences in antenatal care utilization in urban versus rural settings (Navaneetham and Dharmalingam 2002). These conflicting findings may be a result of our use of nationally representative data covering more areas of the country than in past studies. Fourth, our results suggest that women residing in non-EAG States may be more likely to have access to and knowledge about proper contraception and health facilities given that they were more likely to utilize adequate antenatal care and safe delivery compared to those residing in EAG States, similar to findings from other studies (P. K. Singh et al. 2012; L. Singh et al. 2012). However, results also show that women in EAG states are more likely to obtain postnatal care within 24 hours of birth. This might be due to the increased government focus on improving maternal healthcare utilization through various programs in EAG states. For example, from 2004–2008, government spending under the National Rural Health Mission (NRHM) increased the number of women delivering in public health centres in EAG states. Further, an increase in women’s healthcare utilization and a decrease in the cost of delivering in public health facilities was more evident in EAG states compared to non-EAG states (Mohanty and Srivastava 2013). This increase in institutional deliveries in EAG states might have led to a larger proportion of women residing in these areas to avail a postnatal check-up within 24 hours of birth. Finally, consistent with past research, we found that women with more living children in 2005 were less likely to avail adequate pregnancy care (Navaneetham and Dharmalingam 2002). These women may forego extensive care during pregnancy because they have larger families and thus greater resource constraints; they also have more experience with childbirth and therefore may have greater confidence in home delivery and caring for themselves and their babies (Wong et al. 1987; Elo 1992; Bhatia and Cleland 1995; Raghupathy 1996; Navaneetham and Dharmalingam 2002; Santhya et al. 2008; P. K. Singh et al. 2012).

Despite high levels of unintended fertility in India, few studies have evaluated the impact of unintended pregnancies carried to term on maternal healthcare utilization (Jensen and Ahlburg 1999; A. Singh, Chalasani, et al. 2012; Singh et al. 2013; Upadhyay and Srivastava 2016). This is important because enhancing global access to sexual and reproductive healthcare services and incorporating reproductive health into national policies are important targets of the United Nations Sustainable Development Goals (SDG) for 2030 (United Nations Secretary General 2014). Maternal health is a pivotal part of family planning policy initiatives; it is also key to India’s commitments to the SDG of ensuring healthy lives and promoting wellbeing at all ages. Though maternal mortality in India has declined over time, it remains high, especially in EAG States (Sample Registration System 2018). Increasing women’s use of antenatal and postnatal care is likely to bring about improvements in both maternal and child health outcomes (McDonagh 1996; Li et al. 1996; Finger 1997; WHO 2005; Sines et al. 2007; Mattar et al. 2007). Moreover, our results show that unwanted births are very common among women in India—over half of the births (58 per cent) in the sample were unwanted. Thus, strategies for reducing unwanted fertility in India are critical to undertake. Increasing access to and acceptability of effective contraception and abortion care—which is legal in India—would help ensure that women and couples are able to avoid pregnancies they do not want, which will also aid in improving both maternal and child health (see also Singh et al. 2018).

While our study makes significant contributions to understanding differences in women’s utilization of maternal health services during and after pregnancy based on their fertility intentions, there are some limitations. First, our finding that 58 per cent of births in the sample were unwanted is higher than other estimates of unintended fertility from India, which hover around 49 per cent, including unwanted, mistimed, and ambivalently-desired births (Singh et al. 2013; Singh et al. 2018). Due to a lack of data on timing preferences, we are unable to account for ambivalence or mistimed births, which may influence our estimate. This discrepancy may also be because we employ a prospective measure of fertility intentions, and retrospective measures are more sensitive to ex-post revisions toward intendedness (Lightbourne 1985; Bongaarts 1990, 2011; Westoff 1991; Bhushan and Hill 1996; Koenig et al. 2006; Yeatman and Sennott 2015). Additionally, the length of time between the two surveys (7 years), could lead to a misclassification of births since women’s fertility intentions could have changed over the interim period (Westoff and Ryder 1977; Kodzi et al. 2010; Sennott and Yeatman 2012; Yeatman et al. 2013). Second, the measures of the number of antenatal check-ups, the number of tests performed during antenatal check-ups, women’s use of timely postnatal care, and whether the delivery was in an institutional setting for the most recent birth since 2005 were self-reported by women in 2012 and thus could be susceptible to recall bias. This may be especially relevant for the 44 per cent of women in the sample whose most recent birth occurred four or more years earlier. Finally, the variation of PSM that we use as a robustness check could be sensitive to bias when the treatment or outcome model is impacted by confounding unobserved variables (Imbens 2004, 2015; Abadie and Imbens 2006; Kebebe and Shibru 2017). Despite these limitations, we are heartened by the fact that a majority of our results and substantive conclusions are consistent with past research.

Our results linking fertility intentions and maternal health behaviours in India can inform service delivery strategies by highlighting the need to provide women with knowledge about antenatal and postnatal care, and particularly the importance of receiving timely postnatal care for births regardless of where they occur. For example, expanding the use of community health workers, such as Accredited Social Health Activists, could be a fruitful strategy for improving women’s early recognition of pregnancy and knowledge about care options (Vidler et al. 2016). Additionally, when women and their husbands do not want any more children, adequate facilities and services should be available and accessible to them so that they are able to fulfil their intentions and avoid future births (Kost and Lindberg 2015). Providers of maternal health services could improve their counselling services for all pregnant women to encourage them to seek antenatal and postnatal care (Dibaba et al. 2013) aligning with the WHO recommendations. India has several interventions under the umbrella of the National Rural Health Mission (NRHM) aimed at improving maternal and child health in rural regions. For example, the NRHM aims to provide every village in India with trained female community health workers who connect the public health system and the community by providing door-to-door counselling on family planning, the importance of birth spacing, and other aspects of maternal and child health in rural areas. Further, programs such as the Janani Suraksha Yojana (JSY) (launched in 2005) and Janani Shishu Suraksha Karyakram (launched in 2011), encourage women who would deliver at home to instead deliver in institutional settings by giving them incentives such as free delivery (including caesarean deliveries) and free transport from home to health institutions in both rural and urban areas. An impact evaluation of the JSY showed that even though the program significantly increased the likelihood of women obtaining antenatal care and having in-facility births, implementation varied by state (the percentage of women giving birth who obtained cash payments from JSY ranged from less than 5 per cent to 44 per cent), and women who were the least educated and poorest were not always the most likely to obtain cash payments from the scheme (Lim et al. 2010). Thus, improving maternal and child health outcomes in India will also require attention to the structural characteristics that serve as barriers to women equally accessing adequate antenatal and postnatal care, including socioeconomic status, education, and region and area of residence.