Bulletin of the World Health Organization

Socioeconomic inequalities in hospital births in China between 1988 and 2008

Xing Lin Feng a, Ling Xu b, Yan Guo a & Carine Ronsmans c

a. School of Public Health, Peking University Health Science Centre, Xueyuan Road 38, Haidian District, Beijing, 100191, China.
b. Centre for Health Statistics and Information, Ministry of Health, Beijing, China.
c. Department of Infectious Disease Epidemiology, London School of Hygiene and Tropical Medicine, London, England.

Correspondence to Xing Lin Feng (e-mail: fxl@bjmu.edu.cn).

(Submitted: 10 December 2010 – Revised version received: 22 March 2011 – Accepted: 25 March 2011 – Published online: 05 April 2011.)

Bulletin of the World Health Organization 2011;89:432-441. doi: 10.2471/BLT.10.085274


China has made remarkable progress in reducing maternal mortality, and several studies indicate that it may be close to achieving the fifth Millennium Development Goal (MDG 5).13 China’s success has been largely attributed to the strategy of improving hospital delivery: institutional deliveries have risen dramatically, even in the most remote rural areas.2,3 However, the government is now concerned about a widening gap between the rich and the poor in access to health care, including obstetric care.4,5

The boom in China’s economy has reduced poverty overall, but economic reforms have also triggered worry over rising inequalities in access to health care.59 The market reforms introduced in the 1980s have led to a deterioration of preventive health services and a dramatic increase in health-care expenditure.4,1016 The central government decentralized fiscal responsibility to local government, and central government transfers in the health sector have declined.4,1417 As a result, health insurance coverage has decreased sharply and an inability to pay is considered to be the main obstacle for the uptake of maternity care, especially for poor rural households.5,6,18

Over the last fifteen years the government has responded to these concerns in several ways. In 1995 it passed the Law on Maternal and Infant Health Care, which guarantees a woman’s right to institutional delivery. Targets for the proportion of hospital births are imposed by district authorities. In 2000, the government launched a comprehensive safe motherhood programme,19,20 the National Programme to Reduce Maternal Mortality and Eliminate Neonatal Tetanus, targeted at poor rural areas. The programme involved improved infrastructure; enhanced staff training and supervision in township and lower-level hospitals; the establishment of channels for referral to tertiary hospitals equipped to deal with emergency obstetric care; and the provision of subsidies to women and/or hospitals to encourage hospital delivery. By 2008 coverage had increased to more than half of all counties in China. In 2003, the government launched the New Cooperative Medical Scheme (NCMS), which provided further subsidies to encourage hospital delivery.21 The reimbursement policy for maternity care is decentralized and varies by county but, in general, normal deliveries are reimbursed for a flat fee and caesarean sections are reimbursed as a proportion of total expenditure, with a ceiling.18

Whether these policies have reduced the gap between the rich and the poor in access to maternal health care is uncertain. There is no evidence of widening regional disparities in maternal mortality and access to obstetric care over time,2,3 but the evidence on disparities between rich and poor households is inconsistent. Liu et al.5 reported an increase in income-related inequality in hospital delivery rates in rural areas between 1993 and 2003, whereas Long et al.18 found a decrease in the gap in hospital deliveries between the richest and poorest income groups in rural western China between 2002 and 2007. The evidence regarding the impact of the NCMS on the economic burden imposed by health care is also inconclusive.18,2224 According to some studies, the NCMS has not decreased out-of-pocket expenditure;21,24 according to others, it has reduced the out-of-pocket costs of delivery.18,22

The objective of this paper is to examine trends in hospital births in China between 1988 and 2008, especially among the poor. We specifically examine whether trends have changed in the poorest regions and among the poorest households from 2002 onwards, when subsidies for hospital birth were gradually implemented. By focusing our analysis on equity in access to hospital birth, we provide insights into the ways in which the safe motherhood strategy in China may have served the poor.


We used data from four cross-sectional National Health Service Surveys conducted in urban and rural China in 1993, 1998, 2003 and 2008. All surveys were conducted on the same sample of villages, with minor modifications over time, and all used four-stage stratified cluster random sampling to select households. In the first stage, cities/counties were randomly selected from five socioeconomic strata developed by the Ministry of Health.25 This system divides China into socioeconomic regions based on a range of social and economic indicators derived from the 1982 census. The indicators include the employment rate, the percent of the population less than 14 and more than 65 years of age, educational achievements and illiteracy rate, crude birth rate and death rate, infant mortality rate and gross domestic product (GDP) per capita. Each survey sampled a similar number of cities/counties (92, 95, 95 and 94 in the 1993, 1998, 2003 and 2008 surveys, respectively). Within each city/county five districts/townships were sampled, and two villages were surveyed within each district/township. Each survey randomly sampled around 60 households from each village. The total number of households sampled in each survey was 54 984, 56 994, 57 023 and 56 456, respectively.

Locally trained health personnel interviewed household members using a structured questionnaire. All women of reproductive age were questioned about the characteristics of the most recent birth before the survey, including the place of delivery and the number of antenatal visits attended. The recall period was six years for the 1993 survey, two years for the 1998 survey, and five years for the 2003 and 2008 surveys. Data were missing for births in 1994 and 1995.

The questionnaires included a comprehensive list of sociodemographic indicators. Because many of these variables were correlated (data not shown), we selected the following for analysis: socioeconomic region of residence, annual household income, the educational attainment of the mother in the household, access to health insurance, maternal age and parity. Socioeconomic regions were categorized as urban (combining large, medium and small cities) and rural (types I, II, III and IV, this last being the poorest). In each survey households were categorized into five quintiles in accordance with their annual income per capita (gross household income divided by the number of individuals in the household). Thus, the income ranges varied by survey. Per capita income in the poorest quintile, for example, ranged from less than 320 yuan in 1993 to less than 2125 yuan in 2008. Per capita income in the richest quintile, on the other hand, ranged from more than 1200 yuan in 1993 to more than 7500 yuan in 2008. Access to health insurance was defined as any self report of being covered by either a social or a private health insurance plan.

Our analysis was restricted to live births. The outcome of interest was the per cent of births in hospital, including township, county, provincial or national hospitals, including county-level maternal and child care centres. To make our findings representative of the whole country, the data was weighted for the probability of an individual’s inclusion in the sample. We calculated the probability of sample selection at the city or county level by dividing the sampled population by the total population in each city or county, using the census data closest to the date of the survey to calculate the total population.

We used multilevel Poisson regression with allowance for clustering at the city/county and village level and sample stratification to assess crude and adjusted relative risks (RRs) of hospital births over time and in various groups. We chose Poisson regression because the outcome of interest was common and we wanted to compute RRs rather than odds ratios (ORs).26,27 We adjusted the RRs for year, region, household income quintile, mother’s education, health insurance, antenatal visits (4 or more), parity and mother’s age. We modelled year as a linear variable (assuming a constant change per year) and allowed for changes in trend in 1996, when the Law on Maternal and Infant Health Care went into effect, and in 2002, when the major subsidy policies were initiated. We explored yearly trend interactions with socioeconomic region, income quintile, health insurance and parity to determine whether births in hospitals were changing at the same rate over time in all socioeconomic regions and all income, insurance and parity subgroups.


Information on the place of delivery was available for 34 482 live births between 1988 and 2008 (13 727 in 1993; 3981 in 1998; 9006 in 2003 and 7768 in 2008). Trends over time and by region in the proportion of births in hospital are shown in Fig. 1. Hospital births increased from 44.3% of all live births in 1988 to 94.7% in 2008. By 2008, nearly all women in urban and rural China were giving birth in hospital except in the most remote rural region (IV), where only about 60.8% of all live births took place in hospital in 2008. These trends resulted in a reduced gap in hospital births between urban residents and residents of rural region IV (Fig. 1).

Fig. 1. Trends in hospital births by socioeconomic region, China, 1988–2008
Fig. 1. Trends in hospital births by socioeconomic region, China, 1988–2008

Fig. 2 shows trends over time in the proportion of hospital births by income group. The gap between low- and high-income households nearly disappeared: in 2008, 96.1% and 87.7% of high- and low-income households, respectively, gave birth in hospital. The distribution of income groups was not uniform across regions: in urban areas only 2.3% of births were from households having the lowest income, compared with 54.1% in rural region IV. Similarly, 57.6% of urban births were from households with the highest income, compared with 1.9% of births in remote rural region IV. Fig. 3, Fig. 4, Fig. 5, Fig. 6 and Fig. 7 shows trends over time in the proportion of hospital births by region and income group. Data for two-year periods were pooled to improve the precision of the estimates. All income groups in urban and rural regions I, II and III converged towards nearly universal coverage of hospital births. In rural region IV, trends were remarkably similar among the different income groups, with 58.8–62.3% of the women giving birth in hospital in 2008.

Fig. 2. Trends in hospital births by household income group, China, 1988–2008
Fig. 2. Trends in hospital births by household income group, China, 1988–2008
Fig. 3. Trends in hospital births by income group in urban socioeconomic region, China, 1988–2008
Fig. 3. Trends in hospital births by income group in urban socioeconomic region, China, 1988–2008
Fig. 4. Trends in hospital births by income group in rural I socioeconomic region, China, 1988–2008
Fig. 4. Trends in hospital births by income group in rural I socioeconomic region, China, 1988–2008<b> </b>
Fig. 5. Trends in hospital births by income group in rural II socioeconomic region, China, 1988–2008
Fig. 5. Trends in hospital births by income group in rural II socioeconomic region, China, 1988–2008
Fig. 6. Trends in hospital births by income group in rural III socioeconomic region, China, 1988–2008
Fig. 6. Trends in hospital births by income group in rural III socioeconomic region, China, 1988–2008
Fig. 7. Trends in hospital births by income group in rural IV socioeconomic region, China, 1988–2008
Fig. 7. Trends in hospital births by income group in rural IV socioeconomic region, China, 1988–2008

The results of the Poisson regression are shown in Table 1. The proportion of hospital births increased by 6% per year during 1988–1993 (crude RR: 1.06; 95% CI: 1.04–1.09), and this trend was sustained during 1996–2001 (crude RR: 1.06; 95% CI: 1.03–1.09). After 2001, hospital births continued to rise by 4% per year (crude RR: 1.04; 95% CI: 1.03–1.05), although part of this increase was accounted for by sociodemographic change (2% annual trend adjustment). However, the upward trend slowed down significantly between 1996–2001 and 2002–2008 (P-value for change in trend = 0.0076).

The percentage of hospital births differed substantially between regions. Urban women were four times more likely to give birth in hospital than women in rural region IV (crude RR: 3.90; 95% CI: 2.96–5.13) (Table 1). Birth in hospital was much more common among the richest women and those with a college education or higher, but this variation was largely explained by other factors (adjusted RR for lowest versus highest income households: 1.12; 95% CI: 1.03–1.23); adjusted RR for illiterate women versus women with a college education or higher: 1.02; 95% CI: 0.90–1.16). Women with health insurance were more likely to give birth in hospital, but the effect was small (adjusted RR: 1.20; 95% CI: 1.11–1.29). Nearly half (45.9%) of all live births occurred among multiparas (parity of 2 or higher), and such women were less likely to give birth in hospital than women delivering for the first time.

Results of time trends in hospital births by sociodemographic group are shown in Table 2. We pooled the data for 1988–2001 because trends were similar during 1988–1993 and 1996–2001 (data not shown). The yearly increase in hospital births was most pronounced among women living in the poorest region; among the poorest households; among women with insurance and among multiparous women. For example, in rural region IV the average yearly increase in the proportion of hospital births was 21% during 1988–2001 and 10% during 2002–2008. The trends are different across regions (P < 0.0001 for="" 1988–2001;="">P < 0.0001 for="" 2002–2008),="" and="" no="" upward="" trend="" was="" noted="" in="" urban="" areas.="" the="" rise="" in="" hospital="" births="" was="" more="" pronounced="" among="" multiparas="" than="" among="">

The annual increase in hospital births in rural region IV slowed down after 2002 (from 21% per year during 1988–2001 to 10% per year during 2002–2008; P = 0.0031). The annual increase in hospital births among the poorest quintile also slowed down (from 15% during 1988–2001 to 5% per year during 2002–2008; P < 0.0001). hospital="" births="" rose="" dramatically="" among="" women="" with="" health="" insurance="" during="" 1988–2001="" and="" 2002–2008,="" but="" the="" increase="" during="" 2002–2008="" was="" entirely="" accounted="" for="" by="" sociodemographic="">


Using data from four National Health Service Surveys, we reported nationally representative rates of uptake of hospital birth in China between 1988 and 2008. We found that China has made great progress towards ensuring that all women, irrespective of their economic status, give birth in hospital. The gaps between urban regions and poor rural regions and between high- and low-income households are closing fast, and nearly two thirds of women in the most remote rural region now give birth in hospital.

The rising institutional birth rates in rural areas are corroborated by findings from other studies.2,3,11,15,18,28,29 The gap in institutional birth rates between urban and rural regions is narrowing, which suggests that China’s safe motherhood strategy, with its regional focus, is working. However, in recent years the upward trend has been slowing down among those living in rural region IV. In this region, the percentage of institutional births has levelled at around 60% since 2005. This is a cause for concern, particularly since these regions continue to have high maternal mortality.3

The most remote rural region is lagging behind for several reasons. Despite the fact that by 2008 68% of all counties in rural region IV were covered by the national safe motherhood programme19 (this programme targeted counties that had some capacity, even if minimal),19 the most remote counties with poor infrastructure and scarce human resources may have been left behind. Women in remote areas travel long distances to reach a hospital, but smaller township hospitals in these areas may be poorly equipped for deliveries.30,31 The absence of caesarean section and blood transfusion in some of these hospitals may deter women from delivering there.30 Some evidence also suggests that women in the most remote counties, particularly those from ethnic minorities, are reluctant to deliver in hospital because of the discomfort and embarrassment associated with modern birthing practices.3133

According to many studies, inability to pay is one of the main obstacles to the uptake of maternity care in China, especially among poor rural households.10,15,18,34 Our analysis suggests that inequalities between socioeconomic regions are more pronounced than between individual households. It did not show household income to be an important determinant of hospital delivery, since the gap between high- and low-income households has nearly disappeared. The NCMS, which had achieved nearly universal coverage by 2007,24 may have reduced the financial barriers affecting rural households. The NCMS covers only a modest part of the actual cost of institutional deliveries,18,31 and the mean out-of-pocket expenditure reported for institutional births among low-income groups in western rural China is still a substantial burden for the poor in the most remote regions.18,35 Most families now probably recognize the need to give birth in hospital and are willing to bear the additional costs involved.

Lower institutional birth rates among women giving birth for the second or third time have also been found in other studies. This suggests that fear of disapproval or chastisement among multiparas for not complying with government policy on family size may be a deterrent to institutional delivery11,36,37. However, in rural areas and among ethnic minorities the one-child policy is relaxed11,38,39 and higher order births are far from negligible. Encouragingly, the rate at which higher order births are attended in hospital increased by 7% per year between 2002 and 2008.

The data in this study are nationally representative and the sample size is large, but there are limitations. First, births not approved by the family planning system or among rural migrants who were temporarily away from home when the surveys were conducted may have been missed. Migration from rural to urban areas has increased dramatically since 1992, when the sampling frame for the surveys was designed.40 The exclusion of migrants from National Health Service Surveys may have biased the results by affecting trends and income inequalities in institutional births in urban areas. Since migrant women usually belong to the lowest income group and seldom use formal services,41,42 we may have overestimated the proportion of institutional births among the urban poor. Second, rural counties were classified into socioeconomic strata more than 15 years ago and their status could have changed. Nonetheless, the latest National Health Services Survey report suggests that few counties would be re-classified if more recent data were used.43 Third, we used a modified Poisson regression to compute RRs. Logistic regression tends to be the natural choice when the outcome is binary, but the OR overestimates the RR when the outcome is common. To estimate RR directly, binomial regression and Poisson regression with robust standard errors are usually recommended.26,27 We used Poisson regression because the maximum likelihood estimation of the binomial model did not converge.

Institutional deliveries are not a guarantee of good obstetric care. Consequently, one must be cautious when inferring that institutional births have reduced maternal or perinatal mortality.44,45 Institutional births are more conducive to caesarean sections, although the rise in institutional birth rates in China does not explain the rise in caesareans.28 Some studies suggest that obstetric practice is becoming more evidence-informed in county hospitals and higher-level hospitals,46 whereas the quality of care in some rural township hospitals may still be alarmingly poor.30 However, previous research in China has shown that the observed reductions in maternal mortality can be attributed in large measure to increasing institutional birth rates.2,3 Using two different national databases, the authors showed independently that variations in the coverage of hospital deliveries fully explained the declining trends in maternal mortality in China. China’s safe motherhood programme has focused almost exclusively on the promotion of hospital delivery, with doctors rather than midwives attending the deliveries. Midwifery care has been marginalised in China32,33,47 to the extent that very few women give birth with a trained midwife.

China’s sustained commitment to safe motherhood has resulted in impressive improvements in access to institutional birth across all socioeconomic groups, and most Chinese women now give birth in hospital. However, the poorest rural region is still lagging behind and this is causing concern because maternal mortality remains persistently high.3 Ensuring access to hospital delivery in this region is an enormous challenge: roads are poor, distances are long, populations are dispersed, township hospitals are under-resourced and some women may be reluctant to give birth in hospital. A more active mobilization of communities and additional professional and financial resources will be needed to ensure that women in remote regions are transferred to county hospitals where better quality obstetric care can be offered.


We thank the China Medical Board and the Centre for Statistics and Information, Ministry of Health of China, directed by Keqin Rao. We also thank Yanqiu Gao from the Department of Maternal and Child Health, Peking University; Guisheng Liu from the Department of Rural Health, Ministry of Health; and Yaoguang Zhang from the Centre for Statistics and Information, Ministry of Health of China .


Xing Lin Feng is funded by the China Medical Board.

Competing interests:

None declared.