The relationship between internal migration and the likelihood of high-risk pregnancy: Hukou system and high-risk pregnancies in China

This study used inpatient hospital admission data on all women who gave birth between January 1, 2013 and May 31, 2018 at the Shanghai First Maternity and Infant Hospital. These data were used to evaluate maternal high-risk pregnancies. The data contains patient demographics and clinical details on 133,358 live births, with 37% being internal migrants. Our study was approved by the Shanghai First Maternity and Infant Hospital.

High-risk pregnancy status was identified in the medical records and was defined by the National Health Commission of the People’s Republic of China [34]. Specifically, high-risk pregnancy refers to a situation wherein either the mother or the baby were more likely to have health problems during the course of pregnancy, including medical risk, patients with significant medical and surgical disorders, such as chronic hypertension, diabetes, cardiac disease, gastrointestinal disease, etc., and obstetric risk, healthy gravidas with fetuses at increased risk of adverse outcomes, such as multiple gestations, prior intrauterine fetal demise, isoimmunization, etc. This variable was coded as 1 when a mother/fetus had any one of the aforementioned high-risk pregnancies and 0 otherwise [35]. A wide variety of medical and/or obstetric complications are internationally-accepted as high-risk pregnancies during pregnancy, including preeclampsia, gestational diabetes, placenta previa, and fetal problems [36].

Although most internal migrants have rural or urban hukou in their home towns, a host urban hukou remains a crucial determent for internal migrants to access a variety of resources and civil rights in their host city [13]. Guided by our literature review, we define internal migration status by comparing the current type (Shanghai versus not Shanghai) of household registration (‘hukou’). We analyzed mothers’ internal migrant status (Hukou) as a binary variable (1 = Shanghai-born women, 0 = Internal migrants). Sixty-three percent of the sample fell into the Shanghai native-born category (n = 83,872) and thirty-seven percent were internal migrant women (n = 49,486) out of the total sample of 133,358.

In all analyses, we adjusted for demographic and maternal clinical characteristics. Demographic covariates and relevant health variables were also controlled for in order to minimize potential selection bias associated with them, including maternal age at the child’s birth, age squared, gravida—the number of times a woman has been pregnant, birth parity—the number of pregnancies > 20 weeks, ethnicity (1 = Han Chinese; 0 = Ethnic Minority), nationality (1 = China; 0 = others), marriage status (1 = married; 0 = others), severe pregnant, i.e. whether the woman was a critically ill obstetric patient (1 = obstetric critically ill; 0 = not critically ill), cesarean delivery (1 = cesarean delivery; 0 = natural delivery), occupation (1 = employed; 0 = not employed) and insurance status (1 = have health insurance; 0 = without any insurance).

Our analysis consists of two parts. We start with a series of logistic regression models, then, the propensity score matching (PSM) method was used to predict the competing likelihood of having a high-risk condition in pregnancy.

To address potential selection bias, previous studies have used several methods, including instrument variable (IV) analysis and the propensity score matching (PSM) approach [32]. Because it was not easy to find an ideal instrument for migration status, and because the PSM approach works well when dealing with a large data set, such as that used in our study with 133,358 live births over the study period, we selected the PSM approach. The propensity score matching (PSM) approach has been used in several fields to address issues of selectivity bias, heterogeneity and endogeneity [37], and it has been shown to offer estimates that address each potential source of confounding [38–40].

In our study, the outcome variable was the likelihood of a high-risk pregnancy with its determinants estimated through the use of logistic regression methods after controlling for confounding through the use of propensity score methods. The propensity score method was based on a two-stage procedure. In the first stage, the propensity score was estimated in order to identify those characteristics in the dataset that were associated with being a migrant woman. This estimated propensity score was used to match observations in the whole dataset in such a way that individuals with similar values to the identified determining covariates were grouped together in order to assess the independent effect of internal migrant status (Hukou), thereby controlling for potential confounding. In the second stage, the effect of internal migrant status (Hukou) on high-risk pregnancies was estimated using matching techniques based on propensity score estimation. Specifically, we implement each of three different matching methods to assess how sensitive our estimates were to each method: one-on-one matching; caliper matching; and nearest neighbour matching. The parameter we estimated was the average treatment effect (ATE) [41, 42], which in our study context refers to the independent effect of migrant status (Hukou) on the likelihood of a high-risk pregnancy. All estimates were performed using STATA 14.

Table 1 provides descriptive statistics for each of the variables used in the study. Internal migrant status (Hukou), occupation, marriage status, insurance status, nationality, ethnicity, high-risk pregnancy, severe pregnant, and cesarean delivery were coded as binary variables. For example, if the mother was employed, married, had health insurance, not Shanghai-born Han Chinese women, who were high risk, severely pregnant and had a cesarean delivery, then each of these nine binary variables were all one.

The average age of the sample of mothers was 30.7 years, with 37% being internal migrants. Just over half (54%) of the mothers had a high-risk pregnancy. Almost all (94%) of the mothers were employed, 99% were married and 62% had public insurance. In terms of nationality and ethnicity, almost all (99%) mothers were Han Chinese. And almost half (41%) of the women had a cesarean delivery.

Logistic regression was used to estimate the relationship between internal migrant status (Hukou) and the likelihood of having a high-risk pregnancy after controlling for pregnancy status and socio-economic factors, but without imposing propensity score matching. The results are shown in Table 2. The marginal effects demonstrate that there is a high and statistically significant positive effect of internal migrant status (Hukou) on the likelihood of a high-risk pregnancy over the entire study period. We found that infants born to internal migrant women were more likely to be classified as a high-risk pregnancy by 0.038 (column 1, p < 0.001) than their urban counterparts. After controlling for all key explanatory variables, the effect of internal migrant status (Hukou) on the likelihood of a high-risk pregnancy remained significant at 0.050 to 0.060 (columns 3 and 5, p < 0.001). Notably, several covariates were strongly associated with the dependent variable (Table 2). A prior pregnancy, maternal gravitas, severe pregnant had a strong positive effect on the likelihood of having a high-risk pregnancy. Similarly, cesarean delivery was more likely to be associated with a high-risk pregnancy than natural vaginal birth. Besides, medical insurance and ethnicity had a strong negative effect on having a high-risk pregnancy.

To reduce the potential for selection bias, we used propensity score matching (PSM) methods to compare the women who are born in Shanghai to those that moved from other parts of China to Shanghai. Table 3 reports the average treatment effect (ATE) in terms of internal migrant status (Hukou) resulting from three different matching methods, all of which address the issue of confounding. The three methods are (1) one-on-one matching (column 1, row 1); (2) caliper matching (column 1, row 2); and (3) near neighbour matching (column 1, row 3). The estimates of the average treatment effect (ATE) demonstrate that internal migrant status (Hukou) increased the likelihood of high-risk pregnancies by 0.067 (column 1, row 1–3, P < 0.001). The propensity score matching methods yield findings that are almost consistent with the logistic regression results, after controlling for all key explanatory variables, shown in Table 2.

The current literature suggests that migrant mothers tend to be isolated from the rest of society and are less likely to access maternal care which in turn results in their higher likelihood of a high-risk pregnancy, especially for their first child [43, 44]. Besides, under China’s Hukou system, internal migrant mothers are denied urban insurance and often have inferior employment prospects than local residents [10, 11]. Based on these factors, we extended our study by stratifying by birth parity, maternal employment status, and maternal insurance status. We did not stratify by maternal age and marital status since maternal age above 35 years of age was one of the criteria used to define a high-risk pregnancy, and the overwhelming majority of our sample were married (99%).

When we divided the sample into sub-samples by birth parity, maternal employment status, and maternal insurance status, the results were robust. Table 4 reports the results of the effect of internal migrant status (Hukou) on the likelihood of a high-risk pregnancy stratified by birth parity, employment status, and insurance status, using propensity score matching estimates.

We first report the results stratified by birth parity shown in Panel A of Table 4; the results show that internal migrant mothers who give birth for the first time have a higher likelihood by 0.066 (column 1, row 1, P < 0.001) of having high-risk pregnancies than non-migrants having their first child. While internal migrant mothers having their second (or higher order of birth) still had a higher likelihood of having high-risk pregnancies than non-migrants, their higher likelihood was smaller than that for a mother having their first child, 0.027 (column 1, row 2, P < 0.001) compared to 0.066 (column 1, row 1). A possible reason for these results is a lack of access to prenatal health education among internal migrant mothers that may expose them to high-risk pregnancies. However, internal migrant mothers who are having their second (or higher) birth may be sufficiently experienced to make up this deficiency in their lack of access to prenatal care and education thereby lowering their risk of having a high-risk pregnancy relative to those having their first child.

The results of the analysis stratified by employment status are shown in Panel B of Table 4, which demonstrate employed migrant mothers have a greater risk of having high-risk pregnancies by 0.058 (column 1, row 3, P < 0.001) than non-migrants who were employed. The difference persisted but was slightly smaller for those internal migrants that were not employed, who have the risk of having high-risk pregnancies by 0.037 (column 1, row 4, P < 0.001). Internal migrant women usually have poor employment circumstances and low economic status, which together result in worse health outcomes [44]. As such, employed internal migrants tend to represent women in low-paid and inferior-status positions that represent a risk factor for a high-risk pregnancy.

We also extended our results by insurance status. Panel C of Table 4 reports the results stratified by insurance status. Despite having insurance, internal migrants still have a greater likelihood of having a high-risk pregnancy by 0.094 (column 1, row 5, P < 0.001). In the case of those without insurance, there was no statistical difference between migrants and Shanghai-born mothers (column 1, row 6, P = 0.256). Most internal migrants were covered by insurance from their hometown, while China’s system of social health insurance and the Hukou system limited their opportunity to transfer their insurance across regions [45]. Some of them may hold a decent job in their host city and then obtain local insurance. However, insurance coverage still lags behind that for local residents [46]. Internal migrants consequently have a greater risk of having a high-risk pregnancy despite having insurance.

This study assessed the relationship between internal migrant status (Hukou) and the likelihood of having a high-risk pregnancy in the context of China’s Hukou system, using propensity score matching method to reduce for potential self-selection bias into the migration. The present study offers three key findings.

First, we identified a smaller, yet still significant, positive effect of internal migrant status (Hukou) on the likelihood of having a high-risk pregnancy. This result was maintained when we controlled for potential selection bias of migration. Specifically, we found that infants born to internal migrant women were more likely to be classified as a high-risk pregnancy by 0.067. Three alternative methods for controlling for potential selection bias resulted in similar results. This finding is consistent with prior studies in other countries [5, 9, 27, 28, 47, 48]. As mentioned earlier, China’s unique Hukou system denies access to public health services for internal migrant women, which may, in turn, be associated with their higher likelihood of a high-risk pregnancy.

Second, the association between internal migrant status (Hukou) and the occurrence of a high-risk pregnancy differed by maternal birth parity. Internal migrant mothers who were giving birth for the first time had a higher likelihood of having high-risk pregnancies compared to those internal migrant mothers who already had a prior delivery. This finding may be attributed to the fact that mothers who were having a second or subsequent child may have more experience than other mothers, thereby reducing their exposure to a high-risk pregnancy [43, 44].

Third, the difference in high-risk pregnancy persisted by maternal employment status, and maternal insurance status. When we stratified by employment status, those internal migrants who were employed were more likely to have high-risk pregnancies than those who were not employed. This observation may reflect the low pay and inferior employment status afforded to internal migrant mothers who were employed [49]. After stratifying by insurance coverage, internal migrants, even those with insurance, were still having a high-risk pregnancy possibly because the Hukou system may limit their opportunities to transfer their insurance across regions [45].

Our study represents a re-test of the “healthy migrant effect” for pregnant women in China. After addressing potential selection bias, we find that internal migrant women have a greater likelihood of having a high-risk pregnancy, which contrasts with the general observation that migrants report better health status than local residents. Our findings, however, are consistent with existing studies from Shanghai that report worse pregnancy conditions among internal migrants than those reported among their native-born counterparts [11]. This result may be due to the following reasons:

First, China’s Hukou system limits access to public health services, including urban health insurance for internal migrant workers in the host city. As a consequence, internal migrants face significant out-of-pocket costs to their use of such services, limiting their opportunity to receive maternal care that may, in turn, result in adverse reproductive outcomes with high-risk pregnancies being one such outcome [11, 22, 50].

Second, internal migrants, especially women, generally have low status, limited skills, and low-paying jobs, combined with an inability to afford access to medical care, which may further exacerbate disparities in health outcomes [44, 51]. Moreover, the lack of prenatal care awareness may result in an increased likelihood of a high-risk pregnancy [20, 50].

Third, studies also have suggested that internal migrants have a high likelihood of multiple abortions [52], as they are often younger and more likely to have premarital sex. Premarital childbearing is not common in China, therefore these young internal migrant girls are more likely to choose abortion if they were to become pregnant. These factors, when taken together, are important risk factors for a high-risk pregnancy [52].

To the best of our knowledge, this is the first study to use propensity score matching (PSM) methods to address the consequences of migration on the likelihood of having a high-risk pregnancy. Moreover, there very few studies that have focused on adverse maternal health outcomes associated with internal migration, particularly under China’s Hukou system.

In this study, we examine the relationship between internal migrant status (Hukou) and the likelihood of having a high-risk pregnancy in the context of China’s Hukou system. Our findings add to the growing body of evidence that internal migrant mothers in China are more likely to have high-risk pregnancies. First, the study used extensive, detailed, and high-quality medical records data from the Shanghai First Maternity and Infant Hospital. Second, our empirical estimates were based on a clear conceptual framework and addressed the potential for selection bias by using propensity score matching methods. Third, previous studies in China used national-level data, unlike our study that used patient-level health information. Our study was based on inpatient hospital records data, which includes comprehensive maternal medical diagnosis information. The findings remain robust when controlled for a wide range of socio-economic, demographic, and health-related covariates.

The current study has limitations. First, the hospital that was the centre for this study was based in Shanghai, and this hospital may not be representative of all hospitals in China. Consequently, our study findings may lack generalizability. Second, we were unable to directly test various other mechanisms (e.g., maternal psychological attributes, health lifestyle, and life satisfaction) that may account for some of the observed results as such data were not available. Third, women more than 35 years were included in the construct of a high-risk pregnancy. Age itself, was not an indicator of poor health, rather a demographic variable. However, mothers more than 35 years were more likely to have adverse birth outcomes so age does partially represent risk exposure to an adverse outcome [53, 54]. Fourth, we were not able to decompose the indicator for high-risk pregnancies indicator into separate factors. The interpretation of the results is not straightforward. However, high-risk pregnancies refers to a wide variety of medical and/or obstetric complications that are internationally-accepted as high-risk conditions during the course of pregnancy. These conditions are the main causes of maternal and infant mortality. Fifth, the inclusion of only women with live births may lead to bias in the rate of high-risk pregnancies as mothers who lost their babies during the course of pregnancy would not be included. However, because the Chinese government attached great importance to maternal and infant health over the whole course of pregnancy to lower the maternal and infant death rate. This increase in attention has been associated with a decline in infant mortality, such that the neonatal mortality rate (per 1,000 live births) in China reached 3.9 in 2019 [55], and in Shanghai, it reached 2.66 per 1,000 births in 2020 [56]. There is almost no death birth in this study period, therefore the statistic problem of infant death caused by the high-risk pregnancies may not be an issue. Sixth, the internal migrants in our study include internal migrants from other cities, such as Beijing, although they represent a very small subset (0.22%) of the data. Further research is needed to garner a more complete picture of maternal health and its potential interaction with the occurrence of high-risk pregnancies.

This study was approved by the Ethics Committee of the Shanghai First Maternity and Infant Hospital. Date of approving: Jun 13, 2018. Ethical acceptance number: (2018) No. (35). All participants were over 16 years old. Due to the retrospective nature of this analysis and the fact that the data was part of the medical history, informed consent for participants included was waived by the ethics committee on the condition that all data were anonymized. Besides, the dissemination of the study findings does not refer to specific objects but the general source population. The team obtained permission from the Shanghai First Maternity and Infant Hospital to access the data used in this research.

Not applicable.

The authors declare that they have no competing interests.

Our data will not be shared to protect the participants’ anonymity.