Introduction

A number of demographic studies have highlighted the high levels of unintended (mistimed/unwanted) pregnancy in both developed and developing countries. Recent data from the Demographic and Health Surveys indicate that 14–62 % of recent births were unintended [1]. The highest levels of unintended pregnancies are found in the Latin American, Caribbean and South-Southeast Asia regions. Recent estimates from Indian National Family Health Survey-3 conducted in 2005–2006 indicate that 21 % of recent births in India were unintended [2], and approximately 10 % of them were not wanted at all.

Unintended fertility is a concern from both a family planning and a public health perspective. Unintended fertility is often a cause for higher fertility in developing countries including India [3–9]. A substantial body of research exists that highlights the negative consequences of unintended pregnancies for both mothers and children. A number of studies from the developed world have found a significant and positive association between pregnancy intention and delayed initiation of antenatal care and/or decreased number of antenatal care visits [10–21]. However, it is difficult to draw similar inference for the developing world because on one hand there are studies that have reported a positive association between unintended pregnancy and antenatal or delivery care [22, 23] and at the other hand there are also studies which have found mixed results [24–27]. It is important to note that these studies are predominantly based on DHS data.

Evidence on the relationship between pregnancy intendedness and early childhood mortality is even more limited and inconclusive [1]. For example, Montgomery et al. [28, 29] found weak or no effect of unintended fertility on childhood mortality. Whereas, Chalasani et al. [30], using prospective data on pregnancy intendedness found high odds of neonatal and postneonatal mortality among children who were unwanted. A recent study by Singh et al. [27] also revealed higher risk of early childhood mortality among children who were unwanted.

A key issue in investigating the negative consequences of unintended pregnancy is the measurement of unintended pregnancy. A majority of the previous studies rely on retrospective responses from women on whether the indexed child was unwanted or mistimed at the time of birth. Retrospective responses on pregnancy intendedness are likely to suffer from recall bias and ex-post rationalization, i.e., a significant proportion of children who were unwanted or mistimed at the time of pregnancy might be reported as wanted at the time of survey due to various reasons. All these might lead to only ‘most unwanted’ children being reported as unwanted and all others being reported as either mistimed or wanted. These biases collectively might result into systematic under-reporting of unwanted births. A recent study by Koenig et al. [31] which was based in rural India revealed a pronounced tendency for births prospectively classified as unwanted to be retrospectively described as having been wanted or mistimed [31]. The study found substantial difference between the retrospective and prospective assessment of pregnancy intendedness (50 vs. 72 %). The study also provided compelling evidence to infer that the relationship between pregnancy intendedness and negative health outcomes might be biased due to the biases that are inherent in the retrospective responses on pregnancy intendedness.

Given the aforementioned limitations of the previous studies and the inconclusive evidence regarding the relationship between unintended pregnancy and maternal and child health outcomes, we used prospective assessment of pregnancy intendedness to investigate (for rural India) the relationship between pregnancy intendedness and utilization of recommended prenatal care for mothers on one hand and pregnancy intendedness and uptake of a recommended childhood vaccinations for children on the other. We further investigated the relationship between pregnancy intendedness and neonatal and infant mortality.

Data and Methods

Data

Two linked datasets were utilized for this analysis: data from the second round of the Indian National Family Health Survey (NFHS-2) conducted in 1998–1999, and from a prospective follow-up survey of a cohort selected from NFHS-2 carried out in 2002–2003. The NFHS is an Indian version of the Demographic and Health Survey (DHS) and provides consistent and reliable estimates of fertility, mortality, family planning, utilization of maternal and child health care services and other related indicators at the national, state and regional levels. The NFHS-2 (1998–1999) covered nearly 91,196 sample households and 89,199 ever-married women in the 15–49 years age group. The response rates in NFHS-2 were consistently above 90 % for the states selected in our study.

The currently married women in the age-group 15–39 years who were usual residents of households in the states of Bihar, Jharkhand, Maharashtra, and Tamil Nadu at the time of NFHS-2 interviews were followed up in 2002–2003. Bihar and Jharkhand belong to the Northern region (Jharkhand was part of Bihar at the time of NFHS-2), Maharashtra belongs to the Western region whereas Tamil Nadu belongs to the Southern region. These four states together represent the diversity in different demographic, socio-economic and service-delivery indicators in India.

The survey instrument included questions pertaining to respondent’s background characteristics, reproductive behaviour and intentions, quality of family planning, use of family planning methods, an event calendar covering the intervening months between the baseline (NFHS-2) and the follow-up survey, and domestic violence experience [32]. The re-interview rates were high in each of the four states and the re-interviewed and non-reinterviewed samples of women were generally similar in terms of characteristics, indicating that there was no significant selectivity in the re-interviewed sample [32–34]. The details of the follow-up survey can be obtained elsewhere [33].

Outcome Variables

Four outcome variables were included in the analysis. The first outcome variable was the utilization of recommended prenatal care for mother as suggested by the World Health Organization (WHO 2006)—at least four antenatal visits and first antenatal visit in the first trimester of pregnancy [35]. The second outcome variable was the utilization of recommended childhood vaccinations against six vaccine preventable diseases, also referred to as full immunization in India. The recommended childhood vaccinations includes—three doses of DPT, three doses of oral polio vaccine (OPV) and one dose each of measles vaccine and BCG [2]. The other two outcome variables were mortality during first month of life (i.e. neonatal period) and mortality during first 11 months of life (i.e. infant mortality). Deaths during the post-neonatal period (1–11 months of life) were too few, and therefore not estimated. We could not estimate mortality under 5 years of age due to the short birth-history available in the follow-up survey (details of births to women during 1998–1999 and 2002–2003 is only available). Recommended prenatal care, neonatal and infant mortality were based on maternal report. The information on recommended set of childhood vaccinations was based on maternal reports that were compared with the content of the routine immunization card. The discrepancies between the maternal report and the content of the routine immunization card were further probed in the survey.

Exposure Variables

The exposure variable included in the analysis was a prospective measure of pregnancy intendedness in the two linked dataset. The prospective assessment of pregnancy intendedness was made using the procedure devised by Koenig et al. [31]. The responses to questions on fertility preferences in NFHS-2 allowed the assessment of pregnancy intendedness for the inter-survey (between NFHS-2 and the follow-up survey) births. The questions on fertility preferences were:

  • For non-pregnant women:

  • 1. Would you like to have (a/another) child or would you prefer not to have any more children?

  • 2. How long would you like to wait from now before the birth of (a/another) child?

  • For pregnant women:

  • 1. After the child you are expecting now, would you like to have another child, or would you prefer not to have any more children?

  • 2. After the birth of the child you are expecting now, how long would you like to wait before the birth of another child?

Births to women (during the inter-survey period) who stated at the time of NFHS-2 survey that they do not want any more children were classified as unwanted. On the other hand, the births occurring 1 year or more before the preferred time stated by the mother at the time of NFHS-2 interview were classified as mistimed. Births occurring (during the inter-survey period) <1 year earlier, around the same time, or later than the women’s stated preference were all classified as wanted.

Overall, 3,900 non-multiple live births occurred during the inter-survey period. Since the NFHS-2 survey asked only about the mother’s desire for an additional child for the non-pregnant women and additional child after the current pregnancy outcome for the pregnant women, it was not possible to determine the intendedness status for the second or higher-order inter-survey births among the non-pregnant women and first, third or higher-order inter-survey births among the pregnant women. We therefore, had to exclude the second or high-order inter-survey births for the non-pregnant women and first, third or higher-order inter-survey births for the pregnant women from our analysis. In addition, we had to exclude women who were sterilized or believed themselves to be infecund at the time of NFHS-2, and were therefore not asked about their fertility preferences. Further, we also had to exclude births to mothers who reported additional children to be ‘up to GOD’ and births where prospective fertility preferences were not obtained. The result was a sample of 2,347 births for which we could prospectively determine the pregnancy intendedness. Moreover, 239 births in the prospective assessment were wanted-timing unsure, and were thus excluded from the analysis. This resulted in a final sample size of 2,108 births which was used for the analysis.

The other exposure variables included were maternal age (<25 years; 25–34 years; >34 years), woman’s education (none; primary; middle or higher), woman’s autonomy (no autonomy; some autonomy), media exposure (no exposure; partial exposure; full exposure), sex of child (male; female), household standard of living (low; medium; high), and caste (Scheduled Castes (SC)/Scheduled Tribes (ST); Other Backward Classes (OBC); others). Woman’s autonomy was estimated using woman’s responses to three questions about the extent of her decision-making power in her household. Similarly, woman’s media exposure was estimated using woman’s responses on three questions about television, radio, and newspaper use. Women who had no exposure to any of the three media were coded into ‘no exposure’ category, those having exposure to one or two sources of media were coded into ‘partial exposure’ category, and women who had exposure to all the three media were coded into ‘full exposure’ category. We could not include birth order and birth intervals in the analysis pertaining to neonatal and infant mortality because of the unavailability of such information in the follow-up survey. Instead, we controlled for parity (first; others) as done by Koenig et al. [32]. Our categorization of parity into ‘first’ and ‘others’ is also in line with the study by Singh et al. [27] who found that ‘first order’ births were at higher risk of early childhood mortality compared to children of ‘other’ birth orders. For analyzing mortality, we merged the mistimed and unwanted categories of births as well as partial and full exposure categories of media exposure together. We did this in order to adjust the cell frequencies.

Methods

We used bivariate analysis to examine the unadjusted association between pregnancy intendedness and maternal and child health outcomes. Appropriate sample weights were used while estimating bivariate results. We further used binary logistic regression models to examine the adjusted association of pregnancy intendedness with inadequate prenatal care for mother and utilization of inadequate childhood vaccinations for the resultant newborns. In addition we used discrete-time survival models to examine the adjusted association of pregnancy intendedness with neonatal and infant mortality. Survival models were also used to generate the unadjusted estimates of neonatal and infant mortality for the different categories of ‘pregnancy intendedness’. We tested the exposure variables for possible multicollinearity before including them in the multivariate models.

Results

Figure 1 presents the prospective assessment of pregnancy intendedness in rural India. 27 % of the births were unwanted and 12 % of the births were mistimed. For approximately 10 % of the births, the wanted-timing was unsure. Interestingly, only 11 % of the inter-survey births were unwanted and 17 % were mistimed when retrospective responses were used (Results not shown). The characteristics of the 2,108 births shown in Table 1 suggest that about three-fifths of births occurred to women <25 years. Only a little more than 4 % of the births were contributed by women >34 years. 73 % of births occurred to women who had no education and approximately 62 % occurred to women who did not have any autonomy. Media exposure was limited—a little more than two-thirds of the births occurred to women who had no exposure to media. Almost three-fifths of the households in the sample belonged to low standard of living and a little more than half of the households belonged to the Other Backward Clasess (OBC).

Fig. 1
figure 1

Prospective assessment of pregnancy intendedness, rural India, 2002–2003

Full size image
Table 1 Distribution of independent variables and pregnancy intendedness, rural India, 2002–2003 (n = 2,108)
Full size table

Reporting of unwanted births was higher among women aged >25 years, women who had no education, women who had no exposure to mass media, and women who lived in households with low standard of living (Table 1). On the other hand mistimed births were more common among women aged <25 years, women who had middle or higher education, women who had full exposure to mass media, and women who lived in wealthier households.

Figure 2 shows utilization of recommended prenatal care for mother and recommended childhood vaccinations for the resultant newborns by the different categories of pregnancy intendedness. Findings suggest stark variations in utilization of recommended prenatal care and childhood vaccinations by the categories of pregnancy intendedness. For example, 15 % of the wanted births received the recommended prenatal care compared to only 6 % of the unwanted births. Likewise, 39 % of the wanted births received recommended childhood vaccinations compared to only 27 % of the unwanted births. Surprisingly, recommended prenatal care and childhood vaccinations were highest among mistimed births.

Fig. 2
figure 2

Recommended prenatal care for mother and recommended childhood vaccinations for the resultant newborn by pregnancy intendedness, rural India, 2002–2003

Full size image

Unwanted/mistimed births were also disadvantaged in terms of neonatal and infant mortality. The neonatal mortality rate was about 33 per 1,000 live births among the mistimed/unwanted births. This compares with a neonatal mortality rate of only 20 per 1,000 live births among the wanted births. The infant mortality rates among the unwanted/mistimed and wanted births were 49 and 34 per 1,000 live births, respectively (Fig. 3).

Fig. 3
figure 3

Neonatal and infant mortality per 1,000 live births by pregnancy intendedness, rural India, 2002–2003

Full size image

Table 2 presents the results of the adjusted binary logistic regressions. The results adjusted for other socio-economic and demographic characteristics suggest that unwanted births were 2.32 (95 % CI: 1.54–3.48) times as likely as wanted births to receive inadequate prenatal care. Likewise, unwanted births were 1.38 (95 % CI: 1.01–1.87) times as likely as wanted births to receive inadequate childhood vaccinations. But the mistimed births were no different from wanted births to have received inadequate prenatal care or inadequate childhood vaccinations. Births to mothers (1) having primary or higher education, (2) having some autonomy, (3) having exposure to mass media, and (4) living in high standard of living households were significantly less likely to receive inadequate prenatal care or childhood vaccinations.

Table 2 Adjusted odds ratios from logistic regression models for inadequate prenatal care and inadequate childhood vaccinations, rural India, 2002–2003
Full size table

The results of discrete-time survival models for neonatal and infant mortality are presented in Table 3. The findings clearly suggest a disadvantage for the unwanted/mistimed births in terms of child survival. For example, mistimed/unwanted births were 1.83 times more likely than the wanted births to die during the neonatal period. Similarly, mistimed/unwanted births were 1.52 times more likely than wanted births to die during infancy. Surprisingly, there was no association between either household standard of living and neonatal mortality or household standard of living and infant mortality. Sex of the newborn was also not associated with the mortality outcomes.

Table 3 Adjusted odds ratios from survival models for neonatal and infant mortality, rural India, 2002–2003
Full size table

Conclusions

Our findings clearly indicate that unwanted births were disadvantaged in terms of maternal and child health outcomes. Findings reveal that unwanted births were more likely to be accompanied with inadequate prenatal care for mothers and inadequate childhood vaccinations for resultant newborns. Similarly, unwanted births were more likely to die during the neonatal period and during infancy. Interestingly, the sex of the newborn was not associated with mortality outcomes.

Over the last two decades, there has been a spurt in the number of studies examining the negative consequences of unintended pregnancies for maternal and child health. However, most of these studies utilized data from the developed world. Studies examining the relationship between pregnancy intendedness and negative health outcomes for mothers and children in developing countries including India are rather limited, even though the levels of unintended pregnancy are unexpectedly high in these parts of the world. Studies investigating the relationship between pregnancy intendedness and early childhood mortality are even sparser. To date, we could come across only one study that has investigated the relationship between pregnancy intendedness and early childhood mortality for India [27]. Moreover, most of the previous studies on pregnancy intendedness have relied on retrospective responses on pregnancy intendedness collected through a cross-sectional survey. Our study for the first time has used a more rigorous prospective dataset and supports the ongoing debate on the negative consequences of unintended pregnancy for the mothers and the children in a developing country setting, like India. Overall, we found that unwanted births were less likely than wanted births to receive recommended prenatal care and recommended childhood vaccinations. Our findings provide additional support to the studies that also document lower utilization of recommended prenatal care and childhood vaccinations for unwanted children compared to children that were wanted [22, 23, 27]. At the same time, mistimed/unwanted births were more likely to die during the neonatal period and during infancy. Our findings add to the findings of Singh et al. [27] who used a cross-sectional dataset and found a similar relationship. Our findings are also consistent with the findings of Chalasani et al. [30], who using a prospective dataset, also found higher odds of neonatal and postneonatal mortality among children who were unwanted compared to children who were wanted.

The potential limitations of our study must also be highlighted. The follow-up of NFHS-2 was conducted in the year 2002–2003. However, it is the only prospective data that can be used for prospective assessment of unintended pregnancies in India. Also, this data has been effectively used in past epidemiological studies [31, 32, 34]. So we had no other choice than to utilize the follow-up of NFHS-2. Secondly, there are chances of recall bias on the part of mothers regarding prenatal care visits. However, the chances of such errors are least in the carefully designed follow-up survey. Moreover, this is the standard way in which information on prenatal care and childhood vaccinations are collected in the Demographic and Health Surveys. Therefore such biases are not likely to affect our results tremendously. Thirdly, we could not control for the birth order and the birth interval, which are known to affect mortality during early childhood, in the discrete-time survival models due to the unavailability of information in the prospective dataset. However, we controlled for parity. As in Koenig et al. [32], in our analysis also, parity was neither associated with neonatal mortality nor with infant mortality. Another potential limitation of our study is that in the analysis of neonatal and infant mortality, we could not separate the effect of being mistimed from that of being unwanted due to data limitations. So, the effects associated with mistimed/unwanted must be taken as a lower bound for the effect associated with unwanted. While we note that our study has some limitations, the merits of prospective data clearly outweigh those limitations by allowing us to establish temporal association between pregnancy intendedness and negative health outcomes for both mothers and children. Our findings are not only relevant for India but are also relevant for other developing countries that are still in the phase of fertility transition. The countries that are still undergoing fertility transition are likely to face even higher burden of unintended pregnancies.

A key finding of the study was the much lower utilization of recommended prenatal care and recommended childhood vaccinations even in the case of wanted pregnancies—only 15 % of women received the recommended prenatal care and only 39 % of the resultant newborns received the recommended childhood vaccinations. This clearly reflects a major problem regarding access to maternal and child health services across all the groups in rural India. Our findings are consistent with the national averages for rural areas in India [2] and call for greater attention to improving access to maternal and child health services.

In India, unintended pregnancy is not only a reason to worry from the perspective of fertility but is also a cause for concern from the point of view of public health. Therefore, greater attention is required to curb the high levels of unintended pregnancies in India. A recent study has highlighted the role that family planning can play in averting unintended births and in reducing the burden of unintended pregnancy [20]. The authors clearly identify improving access to quality contraception as an important intervention. Our analysis complements the existing literature and argues for re-positioning of family planning (which probably has lost its significance in the years after the International Conference on Population and Development held in Cairo in 1994) in the national reproductive and child health programmes.