1.1 Background

The majority of Indian women (73%) of reproductive age (15–49 years) are married [5]. In the socio-cultural context of India, marriage defines the rights and privileges that women have in society [6,7]. Marriage practices define women’s roles and responsibilities, her place in the family, and her rank and privileges after marriage. In much of India, a woman’s family after marriage is defined as her husband’s family (marital family/home). Depending on a woman’s community and region, most marital arrangements are completely exogamous (marrying outside her village without familial connections). India today remains predominantly a patrilocal society, where women move into their husband’s household after marriage, co-residing with his parents, siblings and their families [8]. Many women are married into households outside of the village, district, or even state of their birth, and thus a married woman often lives far from her natal family [9,10].

Among North Indian Hindus (the region of focus in this paper), this clan-exogamy and patrilocality cause a noticeable break in women’s family environment from natal family to the marital family [7]. This is less pronounced in Muslim families where women tend to retain some of their natal relationships because of a prior relationship with the marital family [7]. However, women’s social status similarly declines as they move into their marital home and enter a new life role as a wife with specific expectations about childbearing [6].

The norms behind fertility and motherhood are contingent on the social environment that women experience after their marriage. The marital family often has control over women’s reproduction, which influences care seeking practices for prenatal, delivery, and postnatal services [6,11]. Women’s autonomy has been found to be positively associated with use of pregnancy care services [12]. For women in rural North India, this could mean that women with higher autonomy are more able to seek more familiar natal environments during last trimester of their pregnancies. Previous research has documented the linkages between cultural and traditional practices and customs related to returning to natal families during pregnancy, childbirth and postnatal period in India [13–17].

The mention of temporary childbirth migration is almost absent from the literature; it appears only as a brief reference within a limited number of studies. A recent systematic review of childbirth beliefs and practices in Asia did not mention temporary childbirth migration [18]. We identified only one study in India by Gawde et al (2016) that quantitatively described temporary childbirth migration [19]. This mixed-methods study interviewed 234 women in Mumbai who had delivered in the last two years, all of whom were internal migrants (migrating within India). Gawde et al. measured how long during pregnancy a woman was “away” at her natal home and whether the delivery occurred in her natal home or current location, comparing recent to longer-term migrants. They found that two-thirds of migrant women returned to their natal home for delivery, with recent migrants spending longer periods of time away; half of recent migrants spent the majority of their pregnancies at their natal home. Qualitatively, reasons for return to the natal home included having more care and support, being expected to work less, and lower health-related costs. However, participants did express some concerns about poorer quality of services at the natal home. While this study is one of the only to describe this phenomenon, it is limited in its focus on rural to urban migrants in one city and has a relatively small sample size.

The remainder of the studies that mention temporary childbirth migration in South Asia are qualitative. The most substantial discussion of this phenomenon is by Raman et al. (2014) in urban Bangalore, India [20]. Interviews with women who delivered in the last two years found that women highly valued the support of their own mothers during pregnancy, childbirth and postpartum. Mothers were seen as the most important source of support, and it was perceived as sad when women could not be with their own mothers during this time. The amount of time women reported staying at their natal homes differed and variability was based on family circumstances. Husbands also visited their wives at their natal homes, and women noted that not only the support of their mothers, but also the support of their friends and extended family in the natal homes was critical. Another study in Tamil Nadu described similar trends of women returning to their natal home during the peripartum period, sometimes for “many months at a time”, adding that this was because the woman’s parents were expected to cover the costs of birth in addition to providing care [14]. Finally, a qualitative study in Bangladesh noted that women preferred to give birth in their natal home, but that financial barriers limited this option for women living in slums [21]. Other studies of traditional practices around pregnancy or delivery mentioned that this practice helped provide mothers sufficient rest, a break from their routine household chores, ward-off potential infection from visitors, or that it was due to views about ‘ritual pollution’ at childbirth [11,22–25]. In Indian society, childbirth is a ‘polluting’ event in which both the mother and the child are considered to be impure or ‘untouchable’ for about 40 days after birth. It is based on the belief that the birth releases the nine months’ of dirt (‘narak’), or vaginal blood, that comes out of the mother’s body at birth which is perceived to cause sickness or ill health [26,27]. One study from a tribal district in Maharashtra found that mother and baby were considered ‘polluted’, and should only be touched by close female relatives including grandmothers, traditional birth attendants and mothers-in-law [28]. Though there is no published statistic of this belief, the concept of pollution has been documented as a major cultural practice during childbirth in other Asian settings including Nepal, Pakistan, and Papua New Guinea [18,29,30].

With the predominantly qualitative literature suggesting that women who temporarily migrate for childbirth do so for the increased support provided at their natal home, temporary childbirth migration has the potential to significantly impact women’s perinatal healthcare experiences through either improving the quality of perinatal care received or creating important discontinuities in perinatal care. Existing literature is limited in addressing this question globally. In Mumbai, Gawde et al. (2016) found that women who returned home reported similar prenatal care use as those who did not; however, women who delivered at their natal home had a 77% reduced odds of facility delivery [19]. Women who returned home; however, were more likely to be visited by a health care worker during the antenatal period. Aside from this, we know little about the impact on health, and thus, an important question about temporary childbirth migration is its impact on health care utilization and outcomes for women and infants.

1.2 Migration and maternal health

Female migration is common in India. Women make up 83% of India’s permanent internal migrants, and 84% of those are marriage migrants [8]. Women also migrate for other reasons such as economic opportunity; however, economic migration only represents 1.1% of women’s migration, consistent with the generally low female labor force participation in India (20.8%) [31–33].

Understanding the health consequences of migration is complicated, as they vary substantially across migrant type, legal status, and other factors [34]. Cross-national immigrants in other settings notably experience a temporary health advantage relative to similar national populations termed the “healthy migrant effect”, despite health care access barriers, environmental and exposures, and behavioral adaptation [35–37]. However, populations represented within this research are often more stable migrant populations, excluding temporary migrants and clandestine workers, whose health status is extremely vulnerable [38]. Internal economic migrants also have heterogeneous experiences and health consequences, with socioeconomic status and social support critically important factors in their health trajectories [35,39]. Migration for females has unique individual opportunities and costs, including specific health vulnerabilities and needs, when compared to men, including around pregnancy and childbirth [40]. This is especially true because peak migration age coincides with peak reproductive age for women, and reproduction requires access to health services.

Much of the research in India suggests that female migrants, especially recent migrants, are less likely to meet key components of the perinatal continuum of care, including meeting the recommended number of antenatal care visits, delivering in a facility, and receiving postnatal care, when compared to non-migrant women [41–43]. One study looking at short-term migrants found similarly reduced use of maternal health services) [44]. However, women migrating from rural to urban areas are generally found to have better outcomes than women “left behind” in rural areas [45]. This could be due to migrant selection, where those that migrate are already more likely to have healthier behaviors or outcomes. However, a migrant’s ability to take advantage of services in new settings may also be based on their ability to assimilate. In a mixed-methods study in urban Lucknow among rural to urban migrant women, migrant women reported poorer respectful care from providers during childbirth than non-migrants [46]. In this study, migrant women explained that lack of availability and experience with services in their rural home communities, combined with social norms around care, contributed to underutilization of these services even after they migrated.

1.3 Care across the perinatal period in India

India has one of the largest government-led Community Health Worker (CHW) program globally with over two million trained CHWs. Two cadres of all-female CHWs work at the community-level and deliver services to pregnant, lactating women and infants: (a) the Accredited Social Health Activist (ASHA) under the National Health Mission and (b) the Anganwadi Worker (AWW) under the national nutrition program, the Integrated Child Development Services (ICDS) [47,48]. ASHAs are honorary volunteer workers who receive a performance-based incentive by the type of services they provide [49,50]. AWWs are paid workers through a network of early childhood development and feeding centres, Anganwadi Centres [47]. ASHAs and AWWs generally work together to deliver essential nutrition and health services to beneficiaries living in their catchment areas (typically, population of 800–1,000). CHW care is intended to reach all residents within their catchment area; however, it is possible that women temporarily migrating for childbirth who are only staying for shorter time periods in the catchment area may be missed, leaving them at risk of care discontinuity.

In 2012, the government introduced a program to strengthen the ICDS policy framework and systems titled the ICDS Systems Strengthening, and Nutrition Improvement Program (ISSNIP) across 162 high undernutrition burden districts in Uttar Pradesh, Madhya Pradesh, Andhra Pradesh, Bihar, Rajasthan, Chhattisgarh, Jharkhand, and Maharashtra to ensure greater focus on children under three years of age. This program consisted of four components, (i) ICDS institutional and systems strengthening; (ii) community mobilization and behavior change communication; (iii) convergent nutrition actions; and (iv) project management, monitoring and evaluation. As part of the ISSNIP initiative, the government launched a digital health intervention, the ICDS-CAS (Common Application Software) for Anganwadi Workers working under the ICDS system. This digital health intervention digitized paper registries maintained by AWWs, digitally entering beneficiary records allowing longitudinal tracking of beneficiaries across different life-stages. It also had reminders for upcoming home visits to beneficiaries’ homes for counselling, auto generation of growth monitoring charts, and tracking immunizations for children and pregnant women.

In this paper, we use a mixed-methods approach to understand trends, risk factors and impacts of temporary childbirth migration in two states of India. First, we describe patterns of women’s displacement across the perinatal period and their access to community health workers while in their natal home. Then we explore socio-demographic predictors of temporary childbirth migration and women’s stated reasons for migrating. Finally, we estimate the impact of being in the natal home, compared to marital home, on prenatal, delivery, and postpartum care. We hypothesize that natal home migration will lead to a disruption in the continuum of care and therefore lower access and utilization of services (Fig 1). An alternate hypothesis in the opposite direction, is that women could receive better care at their natal home. We extend the existing primarily qualitative and descriptive literature by using a large dataset collected from representative samples of women in AWW registers in two states to quantify the impact of Temporary Childbirth Migration on health service use.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Conceptual model of drivers and potential outcomes of temporary childbirth migration.

https://doi.org/10.1371/journal.pone.0292802.g001

2.1 Quantitative component

2.1.2 Measures.

Sociodemographic characteristics captured included age (5-year age categories), educational attainment (illiterate/no formal education; primary (class 1–5); some secondary (class 6–8); secondary (class 9–12); more than secondary)), husband’s occupation (none; agriculture; daily non-agricultural; salaried; other), woman’s working status (outside home vs. not), religion (Hindu vs other), and wealth quintile. Parity was categorized as first birth vs. higher. Sociodemographic characteristics were selected a priori based on previous research in this area and published studies.

Women reported on their return to their natal home during particular time points across the perinatal period: pregnancy (0–3, 4–6, 7, 8, and 9 months), at the time of childbirth, and during the postpartum period (1, 2–5, 6–8, 9–12 months). Women who reported returning to their natal home at any time point were asked why they choose to do this by selecting all that applied from a list of options including: cultural and social norm; get better care, rest of comfort; better or more trustworthy health care; avoid work in husband’s home postpartum; to save money or because husband or family did not want to pay for childbirth; husband asked her to go; or parents asked her to come. Women self-reported on their receipt of perinatal care services relevant to where they were in the pregnancy/postpartum at the time of survey. They were asked the total number of antenatal care visits achieved, the location of childbirth (private hospital, public hospital, other, or home), and whether and how many postnatal visits they received.

We were interested in the impact of being away at the natal home at three points: the last month of pregnancy, at the time of delivery, and in the first month postpartum. For each time period we looked at one outcome related to health care visits. We selected the health care related outcomes to be reflective of guidelines and realize these may not be indicative of actual higher quality care, especially for the case of facility delivery. For pregnancy-related analyses, we made a summary variable of the number of months spent at the natal home in the last trimester of pregnancy to capture the full extent of possible antenatal care. This was operationalized as a categorical variable since there are only 4 possible values. The Indian government officially recommends a minimum of 4 ANC visits, while the WHO recently recommended 8 ANC visits [53]. At least three of these visits are supposed to be in the last month of pregnancy and two in the second to last month. Most of these higher order (4+) visits occur in the later stages of pregnancy, when women are more likely to be in their natal village. As can be seen in Fig 2 most women in India do not meet the Indian guidelines, and only 25% receive more than 4 ANC visits. Since global guidelines recommend more visits, and we know that these visits are most likely to occur at the end of pregnancy, our indicator was constructed as continuous (number of ANC visits). For delivery-related analyses, our primary independent variable was staying at the natal home around the time of delivery. As the government incentive program has pushed the vast majority of deliveries into public facilities [54], and delivering in a private facility is seen as being higher quality, we explored the impact of being at the natal home at time of delivery on a woman delivering in a private facility versus other (public facility or home), although we include delivering in any facility versus home as a secondary analysis. For postpartum-related analyses, we operationalized adequate postpartum visits as receiving a minimum of one postpartum visit because Indian guidelines recommend one postnatal visit in the first 42 days after the mother returns to her home from delivery (post 48 hours) [53].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Distribution of number of antenatal care visits (ANC) received by women in India, using data from the 2016 National Family Health Survey.

https://doi.org/10.1371/journal.pone.0292802.g002

2.1.3 Analysis.

Descriptive analyses of study participants and temporary childbirth migration practices. To characterize our study sample, we first generated descriptive statistics of participant sociodemographic characteristics overall, by state, and migration status (Table 1). We then compared the sociodemographic characteristics of our study participants by state using chi-square tests (Table 1), and due to differences identified and the contextual factors described previously, we compare state-level differences or stratify all subsequent analyses by state. We then characterized the temporary childbirth migration patterns of our study participants through describing the proportion of women who reported being at their natal home during pregnancy (any time during pregnancy, and by category: 0–3, 4–6, 7, 8, 9 months of pregnancy; Table 2), at the time of delivery, and postpartum (any time postpartum, and by category:1, 2–5, 6–8 and 9–12 months postpartum). We also tested for differences in the distribution of temporary childbirth migration and temporary childbirth migration timing by state using chi-square tests (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Sociodemographic characteristics of women 6–12 months postpartum registered at Anganwadi centers in Bihar and Madhya Pradesh, India, 2019.

https://doi.org/10.1371/journal.pone.0292802.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Temporal patterns in temporary childbirth migration among women 6–12 months postpartum in Bihar and Madhya Pradesh, 2019.

https://doi.org/10.1371/journal.pone.0292802.t002

Regression modeling of sociodemographic characteristics and temporary childbirth migration. Next we explored the relationship between participant sociodemographic characteristics and temporary childbirth migration, operationalized as returning to the natal home for at least one month during the following time points: 1) any time in the perinatal period; 2) in pregnancy; 3) for delivery; and 4) postpartum (Table 3 and S1 Table). Each of these four outcomes was modeled separately using mixed-effects logistic regression models with random effects for village to accommodate for our multi-stage sampling structure including clustering of study participants within villages and fixed effects for districts to account for time invariant district specific characteristics. Each model was specified using the following structure: logit(p_TCMij) = β₀ + β₁X_1ij + β₂X_2ij + … + β_kX_{kij +} u_0j + e_0ij where p_TCM represents the probability of temporary childbirth migration at each specific time point, β₀ is our constant, β₁-_k represent the various sociodemographic indicators included in each model (i.e., maternal age group, maternal educational attainment, husband’s occupation, whether the woman works outside the home, caste, first birth, and household wealth quintile: all selected a priori), u_0j is the random effect specific to the jth village and e_0ij is the individual-level error term for individual i in village j. Individual respondents are numbered from i = 1 to N, with each village (j).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Odds of temporary childbirth migration by sociodemographic characteristics among women 6–12 months postpartum, Bihar and Madhya Pradesh, 2019.

https://doi.org/10.1371/journal.pone.0292802.t003

Descriptive analyses of reasons for temporary childbirth migration. Next, we tabulated the reasons that women provide for returning to their natal home among the subset of women who reported any temporary childbirth migration and tested for differences in the proportion of women reporting each of the reasons by state using chi-square tests (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Reasons for temporary childbirth migration among women 6–12 months postpartum who reported returning to their natal home at any point during the perinatal period, Bihar and Madhya Pradesh, 2019.

https://doi.org/10.1371/journal.pone.0292802.t004

Descriptive analyses of perinatal health care outcomes. We described receipt of community-based perinatal health care services provided by either an ASHA or AWW during temporary childbirth migration at the natal home during pregnancy (0–3, 4–6, 7, 8, 9 months of pregnancy; Table 5), at the time of delivery, and postpartum (1, 2–5, 6–8 and 9–12 months postpartum). We then described the proportion of women who achieved each of our three primary perinatal health outcomes: number of ANC visits, delivered in a private health facility, and 1 or more postnatal care visits, and one secondary perinatal health outcome: facility delivery, by state (Table 6a) and migration status (Table 6b) using Wilcoxon rank sum and chi-square tests.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Service receipt from AWW and/or ASHA while at natal home by temporary childbirth migration timing among pregnant and postpartum women in Bihar and Madhya Pradesh, 2019.

https://doi.org/10.1371/journal.pone.0292802.t005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6.

a: Perinatal care achievement among women 6–12 months postpartum by state, 2019. b: Perinatal care achievement among women 6–12 months postpartum by temporary childbirth migration status, total and by state, 2019.

https://doi.org/10.1371/journal.pone.0292802.t006

Regression modeling of temporary childbirth migration and perinatal health outcomes. Finally, we analyzed the impact of temporary childbirth migration on each of our three primary perinatal health outcomes: number of antenatal care visits, private facility delivery, and postpartum health check (Table 7), and one secondary perinatal health outcome: facility delivery (S2 Table), stratified by state. Each of these outcomes was modeled separately using mixed effects generalized structural equation models [55] to simultaneously model the perinatal health outcome and temporary childbirth migration. We selected this approach to accommodate the potential for endogeneity issues whereby certain unobserved factors (e.g., socioeconomic status, health status, etc.) might jointly influence both the likelihood of temporary childbirth migration and the perinatal health outcome, introducing bias into our estimation of this key relationship of interest. To reduce the risk of endogeneity bias, we allowed the error terms in the two equations to be correlated to monitor and mitigate selection bias in those that chose temporary migration. These models accommodated both our multi-stage sampling structure (i.e., through including district as a fixed effect and incorporating random effects at the village level) and our need to combine different model types (i.e., Poisson and logistic) for our varied dependent variable structure [56]. To develop our final models, we first estimated preliminary models evaluating the relationship between sociodemographic characteristics (i.e., maternal age group, maternal educational attainment, husband’s occupation, whether the woman works outside the home, caste, first birth, and household wealth quintile) and each temporary childbirth migration variable of interest to identify which characteristics were conservatively (p<0.1) associated. We then included only the sociodemographic characteristics which met this threshold of statistical significance within each final selection model. Final analyses for the different outcomes followed a similar strategy but differed by dependent variable structure. For example, analyses of temporary childbirth migration dose (i.e., number of months spent in natal home during the third trimester of pregnancy) on the number of ANC visits achieved used Poisson regression models following the two-model structure: 1) log(λ_TCMij) = β₀+ β₁X_1ij + β₂X_2ij + … + β_kX_kij + u_0j + e_0ij where λ_TCMij represents temporary childbirth migration dose, β₀ is our constant, β₁-k represent the various sociodemographic indicators included in each model, u_0j is the random effect specific to the jth village and e_0ij is the individual-level error term for individual i; and 2) log(λ_ANCij) = β₀+ β₁X_1ij + β₂X_2ij + … + β_kX_kij + u_0j + e_0ij where λ_ANCij represents the number of antenatal care visits, β₁ represents the temporary childbirth migration dose, β₂-_k represent the various sociodemographic indicators included and fixed effects for each district, u_0j is the random effect specific to the jth village, and e_0ij is the individual-level error term for individual i. Impact of temporary childbirth migration (i.e., natal home at time of childbirth for delivery analyses and natal home during the first month postpartum for postpartum analyses, respectively) on our two delivery-related outcomes (private versus other facility birth, and facility versus home birth) and achievement of one or more postnatal care visits (PNC) employed logistic regression, following the structure: 1) logit(p_TCMij) = β₀ + β₁X_1ij + … + β_kX_kij + u_0j + e_0ij where p represents temporary childbirth migration, β₁-_k represent the various sociodemographic indicators included and fixed effects for each district, u_0j is the random effect specific to the jth village, and e_0ij is the individual-level error term for individual i; and 2) logit(p_DELVij) = β₀ + β₁X_1ij + β₂X_2ij + … + β_kX_kij + u_0j + e_0ij where p represents the probability of the delivery-related outcome, β₁ represents temporary childbirth migration, β₂-_k represent the various sociodemographic indicators included and district fixed effects, u_0j is the random effect specific to the jth village, and e_0ij is the individual-level error term for individual i. For each model, we monitored for evidence of selection and for models where selection was not identified (i.e., individual random effect variation was 0), we excluded this parameter from the final model presented. No selection was identified within two models: 1) for the outcome of any facility delivery for Bihar (supplementary) and 2) for the outcome of private facility delivery for Madhya Pradesh. We conducted sensitivity analyses modeling district as a fixed-effect versus mixed-effects (Table 7). All analyses were conducted using Stata v.16 [57], and analysis and data files are publicly accessible on GitHub (https://git.ucsf.edu/alison-elayadi/temporary-childbirth-migration.git).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Relationship between temporary childbirth migration during the perinatal period and receipt of perinatal care across three perinatal care periods among women 6–12 months postpartum, Bihar and Madhya Pradesh, 2019.

https://doi.org/10.1371/journal.pone.0292802.t007

2.2 Qualitative component

A qualitative study was conducted in one district in each state in March-April 2018. A total of 32 AWWs and 55 women were interviewed by a team of four Indian female researchers (including the second author (LG)). In Madhya Pradesh, we interviewed 17 AWWs, 13 pregnant women and 17 postpartum women. In Bihar, we interviewed 15 AWWs, 12 pregnant women and 13 postpartum women. The interview guides covered topics related to the intervention in the larger study, as well as questions on knowledge and behaviors related to pregnancy and postpartum infant and maternal health care, and AWW-woman interactions. Interviews were conducted until saturation was reached. Interviews were audio-recorded with permission. The average length of interviews was 90 minutes for AWWs and 30 minutes for women. Interviews were conducted in a private setting either at the homes of the beneficiaries or at the Anganwadi centers for the AWWs.

2.3 Ethics

Study protocols were reviewed and approved by institutional review boards at the University of California, Berkeley (Ref. No. 2016-08-9092), and the India-based Suraksha Independent Ethics Committee (Protocol No. 2016-08-9092). All participants provided informed consent. The interviewer read the informed consent, then the respondent signed on the tablet and was asked to read a sentence providing verbal consent, which was audio recorded on the tablet (“I have understood the purpose of the survey and my rights as the respondent. I agree to participate in this interview.”) The trial is registered at https://doi.org/10.1186/ISRCTN83902145.

3.1 Study sample

Sociodemographic characteristics of our study sample and differences in sociodemographic characteristics by state and by migration status are presented in Table 1. Most study participants were aged 20–24 (47.1% Bihar, 55.9% Madhya Pradesh, no formal education (47.2% Bihar and 32.0% Madhya Pradesh) and works only in the home (88.3% Bihar and 72.0% Madhya Pradesh). The index pregnancy/childbirth was the first birth for 24.3% of participants in Bihar and 32.7% in Madhya Pradesh. All sociodemographic characteristics differed significantly between states and between migration status.

3.2 When do women return to their natal home during the perinatal period?

The distribution of temporary childbirth migration patterns is described across the perinatal period in Table 2. Nearly two-fifths of all women reported having spent at least one month of their most recent perinatal period at their natal home (38.0%). The distribution of when women migrated to their natal homes is bell-shaped, with the largest proportion (25.2%) away for delivery, and slightly fewer (24.7%) away in the 9^th month of pregnancy or during the first month postpartum (23.1%). About a tenth (10.3%) of women report migrating to their natal home in early pregnancy (month 0–3), with steady increases until delivery. Most women report returning to their marital home after one month postpartum, with only 13.3% remaining through months 2–5. More women were away from their marital home in Bihar compared to Madhya Pradesh in most time windows, with no differences after 6 months postpartum.

3.3 Which women return to their natal home, and do the characteristics of women returning to their natal home differ by state?

Older women are less likely to spend any part of the perinatal period in their natal home compared to younger women (Table 3). Compared to women under age 20, women aged 25–29 had about a 40% reduced odds of temporary childbirth migration across most timepoints, with some timepoints not meeting the criteria for statistical significance, and women 30+ had a 60%-70% reduced odds of temporary childbirth migration across timepoints. There was evidence of increased odds of migrating with increasing educational attainment, and this effect was stronger in Madhya Pradesh. Working outside of the home was associated with increased odds of returning to the natal home in Madhya Pradesh, but 30–40% reduced odds in Bihar across timepoints. Being Hindu is associated with reduced odds of returning to the natal home, especially in Madhya Pradesh. The index birth being the woman’s first is associated with increased odds of returning to the natal home, especially in Bihar, but this is driven by being at the natal home during pregnancy, not for delivery or postpartum. There is some evidence that wealth was associated with increased odds of migrating, but results were not consistent across perinatal period and effect sizes were small. Odds of migrating was significantly patterned by district. S1 Table shows the odds of migration at different time points for both states combined.

3.4 Why do women return to their natal home?

Quantitative findings: About half of women who migrated (55%) say they returned to their natal home because they think they will receive better care, rest or treatment from their natal compared to husband’s family (Table 4). Additionally, about 23% report that they thought that the quality of the services were better in their natal home. About 23% say that it was because their parents asked them to come. Fewer (16.7%) report that it is the culture/tradition (with more saying this in Madhya Pradesh), 11.2% because their husband’s family asked them to go, 9.5% to avoid work in their husband’s home (again more said this in Madhya Pradesh) and 4.3% because their husband and his family did not want to pay for services.

Qualitative findings: Mirroring the quantitative findings, women report usually going to their natal home in the last trimester of pregnancy, any time between months 7–9, and usually stayed there through the first 3 months after delivery. Most women describe the decision about whether or not she would return to her natal home as being made by their husband or in-laws. However, women often also mention a preference for returning to the natal home for delivery as they anticipated receiving better care (social support, attention) or being able to delay having the next birth for longer (especially if she had a daughter, which would lead to pressure to rapidly become pregnant again to bear a son). Women also discuss not having to do as much work at their natal home as in their marital home in the postpartum period. These perspectives are expressed in the quotations below:

In my maternal place, mother takes good care (of me), all necessary food items are there. (There is) no comparison of mother; mother in law is just mother in law, and she can’t take place of real mother.

(Age 19, 1 birth, currently pregnant, 10^th grade education, Madhya Pradesh)

Because there was no one to take care of me…. My husband is alone and he use to go to sell vegetables and I have my two year old daughter also, so my mother called me [asked me to come]

(Age 26, 2 children, 12-24m postpartum, 5^th grade education, Madhya Pradesh)

3.5 What is the impact of returning to women’s natal homes during the perinatal period on care access?

Quantitative findings: Among the full sample, 44% of women who returned to their natal home report that they saw either an AWW, ASHA or both AWW and ASHA while they were in their natal home. This is higher in Madhya Pradesh than in Bihar for every single time period, with a range of 53–64% of women reporting this in Madhya Pradesh and 18–35% in Bihar (Table 5).

Antenatal Care: Women in Bihar received a median of 2 ANC visits (Interquartile Range (IQR) 1–3) and women in Madhya Pradesh received significantly more visits, with a mean of 4 (IQR 2–6) (Table 6a). There was no significant difference in the median number of ANC visits by migration status, with non-migrating women having a mean of 3 (IQR 2–4) and migrating women a median of 3 (IQR 2–5) (Table 6b). In multivariable joint models incorporating the potential for selection into migration based on sociodemographic characteristics, women who spent more months in their natal home in the last trimester of pregnancy had a significantly higher ANC visit rate in Bihar but not in Madhya Pradesh (Table 7). Each month spent at the natal home in the third trimester of pregnancy was associated with a 5% and higher ANC rate in Bihar (Incidence Rate Ratio (IRR) 1.05, 95% CI 1.02–1.08). Women’s education and household wealth were consistently positively associated with more ANC visits in both states, and ANC visits were patterned by district.

Delivery: A higher percentage of women who returned to their natal home (13.8%) delivered in a private facility, compared to those who did not (10.6%) (Table 6b). More women in Bihar delivered in a private facility compared to in Madhya Pradesh (11.1% compared to 5.1%) (Table 6a). However, in multivariable models accounting for selection into migration based on sociodemographic characteristics, women who were at their natal home at the time of delivery were no more likely to deliver in a private facility versus a public facility in both states (Table 7). Findings were consistent in supplementary analyses of home birth versus any facility which found that women delivering in their natal homes were also no more likely to deliver in any facility (public or private) compared to home births (S2 Table). Education and wealth status were again consistently positively associated with both increased odds of delivering in a private vs. public facility and delivering at any health facility versus home. Again, private facility and any facility outcomes were patterned by district.

Postnatal care: About 50% of all women received 1 or more PNC visit, with no differences by migration status (Table 6b). Women in Madhya Pradesh were more likely to have one or more PNC visit compared to women in Bihar (48.7% vs. 25.8%) (Table 6a). Being in the natal home in the month immediately following delivery was not associated with PNC visits in either state in multivariable models accounting for potential sociodemographic selection effects (Table 7). Working outside of the home and wealth status were associated with higher odds of PNC visits, as was working outside the home, but in Bihar only. PNC visits were patterned by district.

Qualitative Findings: Anganwadi workers (AWWs) report that temporary childbirth migration is a substantial challenge for providing their care. They report that there is no way to track women who leave for their mother’s village or identify women who come to their village because it was where their parents lived. A few AWWs also mentioned advancing the baby shower celebrations (part of the AWW services) for pregnant women before they leave for their natal homes, as a way of addressing childbirth migration. Most are unable to provide the appropriate counselling to women throughout the perinatal period, although some are able to work around this issue by through calling women on the phone or offering services earlier in pregnancy.

“When I returned after four months (from my parents’ home), the [AWW] came to know so she came after one month to make the card…. Yes. she could know only if I had informed her. Otherwise how would she know if I have returned or not.”

(Beneficiary, age 21, 3rd birth, 6–12 month postpartum, 12^th grade education, Bihar)

We do baby shower between the fifth month and the seventh months. Many times some [pregnant women] are left because she goes to her parent’s place in the seventh month… In it (the baby shower) we (AWW) give all the things given in godh bharai (baby shower) like bangles, bracelets, nariyal (dry coconut) etc. That is why we do it in the fifth month and give all the information so that no one is left out.

(AWW, Age 45, Undergraduate education, Madhya Pradesh)

AWWs also note that they are unable to enroll women temporarily at their natal homes in supplementary food programs provided by the government to pregnant women and newborns because they are not officially registered in their natal villages.

How would I give them the take home ration (supplementary nutrition), already I have to choose from the many daughters-in-law in our village. This selection only creates a lot of problem, I can’t even think of giving it to any daughter of the village (daughter of the village implies the women staying at their natal home)….

(AWW, Age 40 years, Undergraduate education, Bihar)

4.1 Explanations for state-level differences

One of the main findings of this analysis is that the pattern of maternal childbirth migration differs by states of India (even states that are relatively close to each other geographically), perhaps highlighting nuanced cultural practices during childbirth. Slightly more women returned to their natal home in Bihar compared to Madhya Pradesh; however, the difference was not statistically significant. While about 40% of women still reported some care from a CHW while at the natal home, the proportion was much smaller in Bihar compared to Madhya Pradesh. This suggests that not only does the prevalence of temporary childbirth migration differ by state, but what happens when women return to their natal home (in terms of connection to and continuity of care from their CHWs) differs. Gawde et al. (2016) in Mumbai, Maharashtra state, found that about two thirds of women returned home, however, this was a much smaller sample of women, all of who were migrants in an urban area [19]. Thus, collecting data from different regions among a diverse populations is an important next step for understanding this trend further, or within a specific population of interest.

Furthermore, we only found associations between being at the natal home and better health care use (for ANC) in Bihar, and not in Madhya Pradesh. Both Madhya Pradesh and Bihar fare poor on many maternal and child health indictors, as evidenced by both having a high burden of undernutrition indicated by high under-five mortality and anemia among pregnant and lactating women; however, compared to Madhya Pradesh, Bihar has worse indicators on health and nutrition program coverage and services including the Integrated Child Development Services (ICDS) [5]. These differing contexts of ICDS governance and implementation capabilities could help explain our findings. Bihar generally performs worse in most maternal, neonatal and child health indicators and services, and thus the added benefit of care and support (including emotional and financial) from the woman’s natal home may make a bigger impact in Bihar [5]. On the other hand, in Madhya Pradesh, ICDS and maternal and neonatal health services are better, thus, women may be able to weather the disruption of the continuum of care period better. For example, 5.7% of women in Madhya Pradesh compared to 14.4% in Bihar received 4 ANC visits, 80% women in Madhya Pradesh deliver in facilities compared to 63.8% women in Bihar, and 17.5% of women in Madhya Pradesh compared to 10.8% in Bihar receive postnatal check-ups with 2 days [5]. Lower antenatal care and delivery uptake in Bihar is reflective of inadequate health infrastructure in Bihar, and therefore added support from natal families will make more of a difference. A recent Government of India’s report documented systemic and structural deficiencies in Bihar’s health system, including shortage of blood banks, shortage of beds, inadequate administration of tetanus toxoid injections to pregnant women, among other indicators [61]. Another explanation is that CHWs in Madhya Pradesh might more effectively connect with women who migrate, perhaps because it’s a better functioning health care system broadly, making differences between the groups minimal in terms of coverage in Madhya Pradesh. Future research to understand the impacts of temporary childbirth migration should incorporate these multi-level factors.

4.2 Implications for research

Current data collection for household surveys in India including for the National Family Health Survey, do not account for temporary childbirth migration (specifically: if, where, when, duration or the impact of temporary childbirth migration is not collected). For household surveys collecting retrospective data on care received in the perinatal period and outcomes, this could lead us to misinterpret findings by geographic location. For example, a survey may report receipt of care by women in a district, but it does not mean that all these women received the care from CHWs of the same districts throughout her ANC and PNC.

Our analysis has shown that state level context mattered, with patterns and impacts differing between Bihar and Madhya Pradesh, so district or lower-level contexts should also matter. Indeed, our fixed-effects analyses identified some differences in migration behaviors and outcomes within district-specific comparisons which suggests that future research should further explore the role of district-specific contextual variables. About one-third of women had their natal home in a different district and one-third in a different block. Thus, for studies identifying women through registration lists from an Anganwadi center or a primary healthcare center, women not registered locally may be missed, and for studies where women are recruited at home, women not home when a data collector visits them where they usually live and are, presumably, not registered. Furthermore, specific factors may be associated with both temporary childbirth migration and adverse maternal and neonatal outcomes, resulting in biased findings without inclusion of these participants.

Temporary childbirth migration also causes problems for data collected from health facilities. If women are delivering at a different facility from where they received ANC, she might be recorded for the birth, and, given no previous health record at that facility, it might be assumed that she did not receive ANC, when in fact, she did, just at another location. While women do have a card that records their medical information, they may not always carry this or it may miss some information. This could be affecting our interpretation of the association between receiving care and delivery outcomes (complications, etc.). Finally, from census data, if a “household” is defined by people who have been sleeping there for a set amount of time, women may be left off, or double counted, or, potentially misclassified as being another type of migrant.

4.3 Strengths and limitations

This study has several strengths, including a large primary dataset collected from two Indian states which specifically asked about a variety of practices related to temporary childbirth migration, representing a more diverse sample of women and more comprehensive data collected than previous studies. Furthermore, we captured both quantitative and qualitative data from multiple stakeholders on this phenomenon. Our study also makes important contributions to the literature. Many studies that use Demographic Health Survey datasets focus on child health outcomes and relate them to exposure to pollution or shocks during month of birth but use residential location as the place where these shocks were observed (which may not be correct). However, data from this study suggest that this may not be an accurate assumption in all cases. It should be noted that a few studies have considered month of birth along with residential location [62,63].

However, several limitations do exist. First, we do not have a full sample of women who potentially migrate, as some women may have still been at their natal homes at the time of interview. We reduced the potential impact of this through limiting the sample to women who were at least 6 months postpartum; however, it is possible that certain women, who may be different systematically, were missing. Additionally, this restriction of the sample also reduced the sample size. Second, we only include data from two states which are considered among severely under-resourced states in India, and our findings may not be generalizable outside of these states and populations. Additionally, our sample, which overrepresented rural and more disadvantaged populations, likely underestimates this trend overall since we find that more educated and richer women are more likely to migrate. Our data also come from women who are registered at Anganwadi Centres, so the results are only generalizable to women from these districts who are registered in the ICDS system. The NFHS-4 in India, which was collected around the time of this study, found that 65% of women who gave birth in the last 5 years in Bihar and 81% in Madhya Pradesh were registered [5]. Thus, these findings can only be interpreted as describing temporary childbirth migration and associated impacts on care among a population of women who are already in the health care system during their pregnancies. Women who are never registered in pregnancy may be unlikely to receive care at all, and thus might face no impacts of migrating during pregnancy (since they were never seeking care). On the other hand, women who do not register in their husband’s home may migrate and somehow seek care in their natal home. More research is needed on this population. Data on health outcomes would add depth to our understanding of this phenomenon and should be collected by future studies. Given the hierarchical nature of our study sample and potential for endogeneity and selection bias, the models used for our primary analyses of the impact of temporary childbirth migration were complex, resulting in lower levels of precision around our effect estimates. As a result, our results may be conservative in identifying a true impact of migration on health outcomes. The qualitative study also purposively sampled women for understanding interaction with CHWs and did not specifically seek to explore the reasons and factors for childbirth migration. Finally, the retrospective nature of the data leads to the possibility of recall bias.