Bulletin of the World Health Organization

Time trends and regional differences in maternal mortality in China from 2000 to 2005

Gao Yanqiu a, Carine Ronsmans b & An Lin a

a. Peking University Health Science Center, 38 Xueyuan Road, Haidian District, 100083 Beijing, China.
b. London School of Hygiene and Tropical Medicine, London, England.

Correspondence to An Lin (e-mail: anlin@bjmu.edu.cn).

(Submitted: 17 October 2008 – Revised version received: 23 March 2009 – Accepted: 07 April 2009 – Published online: 25 August 2009.)

Bulletin of the World Health Organization 2009;87:913-920. doi: 10.2471/BLT.08.060426


Twenty years into the Safe Motherhood Initiative, countries continue to struggle to reduce the high burden of maternal death. The fifth Millennium Development Goal (MDG-5) has set a target of reducing maternal mortality by 75% between 1990 and 2015, but progress has been slow.1,2 Effective technical interventions exist, but their application requires considerable skill. The training, deployment and retention of skilled health personnel remain huge challenges, and inequities, insufficient financial resources and lack of political will continue to limit progress.3,4

China represents one of the few success stories in maternal health. The maternal mortality ratio (MMR), estimated at 1500 deaths per 100 000 live births in the 1950s,3,5 decreased to an estimated 88 deaths per 100 000 live births in 1990.68 The factors explaining this decline are complex, but the successful expansion of rural health services, with an effective referral system from villages to township and county hospitals, is thought to have been a major factor.911 During the past 25 years China has undergone unprecedented economic growth, and the health reforms introduced concurrently have raised concerns over rising inequalities.12,13 There is ample evidence of inequalities between regions in access to obstetric care and maternal survival, but whether gaps increase over time is unknown.7,1315

As in other countries, in China the accuracy of maternal mortality estimates is uncertain. The recent Countdown, which tracks coverage of child and maternal health indicators in 68 countries, including China, was unable to track progress towards MDG-5 because of the huge margin of uncertainty in maternal mortality estimates.16 China’s maternal mortality estimates are typically obtained from a sentinel surveillance system covering a sample of 37 urban and 79 rural sites.17 The generalizability of these data has been called into question, and discrepancies have been noted in the classification of the causes of death.7

In this study we used a different source of data to examine trends and variations in the MMR in China between 2000 and 2005. The National Maternal and Child Health Routine Reporting System, established in the 1980s, covers the entire population of China.18 We report trends in maternal mortality by province and region, and explore the extent to which the observed trends are explained by changes in the proportion of institutional births, the crude birth rate and economic growth over time.



We obtained data on live births and maternal deaths in each province between 2000 and 2005 from the National Maternal and Child Health Routine Reporting System, which falls under the responsibility of the Ministry of Health.13,18 The system differs somewhat in urban and rural areas. In urban areas, all pregnancies are registered and community doctors keep a log of all pregnancy outcomes. When the pregnancy outcome is not known, they call the woman’s home to update the information. Once a month they also visit obstetric and emergency departments in their catchment area and check death certificates at the police departments to further identify maternal deaths. Community doctors send monthly reports to sub-district health managers, who forward summary reports to district health managers twice a year. In rural areas, village doctors use their extensive community networks to identify births and deaths within their catchment area. Data are forwarded monthly to township hospitals and twice a year to the county health department. Maternal deaths are defined as the death of a woman while pregnant or within 42 days of pregnancy termination, irrespective of pregnancy duration or termination method, excluding deaths from intentional and unintentional injuries.19

We also obtained data, by year and province, on the percentage of live births in institutions18 and on per capita gross domestic product (GDP) and crude birth rate from the National Statistical Yearbooks. The per capita GDP product was converted to United States dollars (US$) at the exchange rate of 6.83 yuan to 1 US$ (14 August 2009). Institutional births are defined as births in township, county, provincial or national hospitals. Nominal GDP values were converted to constant currency units using the year 2000 as a base.

Statistical analysis

We assigned each of the 31 provinces, autonomous regions and municipalities to one of three economic regions, using the Chinese government’s definition of the eastern (Beijing, Tianjin, Liaoning, Shanghai, Shandong, Jiangsu, Zhejiang, Fujian and Guangdong), central (Hebei, Shanxi, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei, Hunan and Hainan) and western region (Inner Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia and Xinjiang). We used box plots20 to model the MMRs, the percentage of live births in institutions, the per capita GDP and the crude birth rate in each province by year and region. Boxes represented the interquartile range (IQR) (from the 25th to the 75th percentile), and any data which lay more than 1.5-fold the IQR below the first quartile or 1.5-fold the IQR above the third quartile were considered outliers.

We examined time trends in maternal mortality with Poisson regression. We first computed crude annual relative risks (RRs) with 95% confidence intervals (CI), using year as a continuous variable and adjusting for clustering at the province level, and tested whether time trends were similar across the three regions by introducing an interaction between year and region. Second, we reported the relative and absolute difference in MMR between regions in 2000 and 2005, using Poisson confidence intervals. Third, we calculated crude annual trends in mortality within each province. Finally, we performed multivariate analysis by progressively adding region, the percentage of births in institutions, the GDP per capita and the crude birth rate to the model in order to predict annual trends (adjusting for clustering within province). Stata version 10.0 software (Stata Corp., College Station, TX, United States of America) was used for statistical analyses.


Between 2000 and 2005, the registration system recorded 30 672 maternal deaths and 64 780 153 live births, resulting in an overall MMR of 47.3 deaths per 100 000 live births. As shown in Table 1 and Fig. 1, MMRs in the western region were on average four-fold higher than in the eastern region (crude RR: 3.81; 95% CI: 3.00–4.83). Similarly, MMRs in the central region were twice as high as in the eastern region (crude RR: 1.87; 95% CI: 1.54–2.28). There was little variation in mortality across provinces in the central and eastern regions, but mortality varied substantially within the western region. The six outliers in the western region were all in Tibet.

Fig. 1. Box plota of MMR by region and year, China, 2000–2005
Fig. 1. Box plot<sup>a</sup> of MMR by region and year, China, 2000–2005
IQR, interquartile range; MMR, maternal mortality ratio.a Any data which lie more than 1.5-fold the IQR below the first quartile or more than 1.5-fold the IQR above the third quartile are considered outliers.

Maternal mortality declined by an average of 5% per year (crude RR: 0.95: 95% CI: 0.94–0.97). There was no interaction between region and year (P = 0.2311). Mortality declined by 5% per year in the eastern region (crude RR: 0.95; 95% CI: 0.92–0.97), by 5% per year in the central region (crude RR: 0.95; 95% CI: 0.94–0.96), and by 4% per year in the western region (crude RR: 0.96; 95% CI: 0.94–0.98). There was no evidence of a widening gap in maternal mortality between regions between 2000 and 2005 (Table 2). Impressively, the absolute difference in MMR between the western and eastern regions declined from 65.4 deaths per 100 000 live births in 2000 to 49.4 per 100 000 live births in 2005.

The MMR declined significantly in four of nine eastern provinces, in six of 10 central provinces and in eight of 12 western provinces (Table 3). Mortality decreased significantly in all five provinces with an MMR greater than 100 per 100 000 in 2000. In Tibet, mortality declined from 466.9 deaths per 100 000 live births in 2000 to 290.3 deaths per 100 000 live births in 2005 (crude RR: 0.93; 95% CI: 0.89–0.98).

The percentage of births in institutions ranged from 20.1% in Tibet in 2000 to 99.7% in Tianjin in 2004, with a median of 81.9% (Fig. 2). In the eastern region, 98.5% (range: 78.1–99.7%) of women gave birth in health institutions, compared with 83.3% (range: 57.3–94.7%) in the central region and 66.6% (range 20.1–90.9%) in the western region. The extremes at the low end of the distribution were in Tibet (increasing from 20.1% in 2000 to 34.1% in 2005) and Guizhou (increasing from 25.8% in 2000 to 48.7% in 2005). The crude association between institutional births and maternal mortality was strong: maternal mortality was four times higher in provinces where fewer than 60% of births took place in institutions compared with provinces where more than 90% of women gave birth in an institution (crude RR: 4.51; 95% CI: 3.60–5.63) (Table 1).

Fig. 2. Box plota of the percentage of institutional births by region and year, China, 2000–2005
Fig. 2. Box plot<sup>a</sup> of the percentage of institutional births by region and year, China, 2000–2005
IQR, interquartile range.a Any data which lie more than 1.5-fold the IQR below the first quartile or more than 1.5-fold the IQR above the third quartile are considered outliers.

Per capita GDP in 2005 increased dramatically, from a median of US$ 860 in 2000 to US$ 1564 (Fig. 3). Per capita GDP was much higher in the eastern region (median: US$ 2575; range: 1399–7707) than in the central (median: US$ 1165; range 710–2023) or western regions (median: US$ 911; range: 390–2235). There was a strong association between per capita GDP and maternal mortality (Table 1).

Fig. 3. Box plota of GDP by region and year, China, 2000–2005
Fig. 3. Box plot<sup>a</sup> of GDP by region and year, China, 2000–2005
GDP, gross domestic product; IQR, interquartile range; US$, United States dollars.a Any data which lie more than 1.5-fold the IQR below the first quartile or more than 1.5-fold the IQR above the third quartile are considered outliers.

The crude birth rate varied between 4.85 and 13.95 per 1000 population in the Eastern region, between 7.25 and 15.44 per 1000 population in the Central region and between 9.05 and 19.05 per 1000 population in the western region (Fig. 4).

Fig. 4. Box plota of crude birth rate by region and year, China, 2000–2005
Fig. 4. Box plot<sup>a</sup> of crude birth rate by region and year, China, 2000–2005
IQR, interquartile range.a Any data which lie more than 1.5-fold the IQR below the first quartile or more than 1.5-fold the IQR above the third quartile are considered outliers.

The trends in time were fully explained by changes in the percentage of institutional births (Table 1). After adjustment for region and variation in institutional births, the RR in maternal mortality for one year relative to the previous year was 1.01 (95% CI: 0.98–1.04). After adjustment for all factors in the multivariate model, there was some residual variation in maternal mortality according to region, the percentage of births in health institutions and per capita GDP (Table 1).


China has made great strides in reducing maternal mortality. If the impressive annual decline of 5% persists, maternal mortality will decrease by 70% during the next 25 years. Encouragingly, progress is seen in all regions, even in the most remote and poorest provinces, and there is no evidence of a widening gap between better-off and economically more deprived regions.

What explains China’s success? China has made substantial efforts to improve access to hospitals and emergency obstetric care.9,10,14 Gains were possible in part because of the hierarchical health system, which provides effective referral for high-risk patients and an extensive supervision system that facilitates training and contact between health facilities, even in remote areas.11,15 Within this system, the promotion and provision of safe delivery services continue to be key priorities.2123 The “Decreasing Project” initiated by the government in 12 western rural provinces in 2000 aimed to decrease maternal mortality and eliminate neonatal tetanus.24 Maternal emergency referral systems were strengthened, the quality of services was improved, and targets for increasing institutional birth rates and improving service quality were imposed. Our data show that institutional birth rates have risen substantially, particularly in the western region, where institutional birth rates increased from a median of 58% in 2000 to a median of 74% in 2005. This increase undoubtedly accounts for a large proportion of the observed decrease in mortality. Other factors may also have been important, including an increase in household income leading in turn to increased affordability and demand for hospital delivery. The role of the one-child policy is difficult to ascertain. It is reasonable to assume that families invest more in care for pregnancy and birth when such events occur only once or twice in their lives. However, the family planning policy is unlikely to explain the trends in maternal mortality during the short period studied here. The crude birth rate did not change between 2000 and 2005, and birth rate was not associated with maternal mortality in our adjusted model.

Even after adjustment for the increase in institutional births, variations in maternal mortality persisted across regions and per capita GDP. The regional differences are particularly notable, since the MMR in the western region remained twice as high as in the eastern region. Women in the western region were more likely to give birth in smaller township hospitals, which may not be sufficiently equipped to deal with obstetric emergencies.21,22 When access to institutional birth attendance is low, women may also seek professional care too late for a midwife or doctor to be able to save their lives.25

The improved access to hospital care for birth goes hand in hand with more technical interventions in normal births. The use of antenatal ultrasound scans is common, and the rapidly increasing Caesarean delivery rates reflect, in part, women’s increasing preference to give birth by this route.15,26,27 The population-based Caesarean delivery rate increased from 6.7% in 1993 to 20.6% in 2003.28 However, the shift towards interventionist births may not be economically sustainable,15 and Caesarean delivery may cause avoidable morbidity and mortality.29 The government is aware of the increased medicalization of childbirth, but efforts to reverse these trends are clearly needed.30

The consequences of China’s health care reform on growing inequities in survival and access to care have received much attention.7,13,15 State expenditure as a proportion of total health expenditure has decreased, out-of-pocket payments have increased, and efforts to revive health insurance programmes have had variable success.14,15 Our data do not support the hypothesis that the gap between rich and poor regions has widened. The MMR has declined at a similar rate in all three regions, the absolute difference between regions has become smaller, and provinces with the highest burden of mortality have made the greatest absolute progress. We could not assess mortality in rural and urban areas or for isolated ethnic minorities or migrant populations, but the data from the most remote provinces are certainly encouraging.

A critical assessment of the accuracy of the vital registration data is warranted. Even though the Maternal and Child Routine Reporting System was established in the 1980s, we used data from 2000 onwards because the system was substantially strengthened in 2000 and in later years. The levels of maternal mortality estimated here are remarkably close to those reported by the sample surveillance system.6,8,17 This is encouraging, because the latter covers only a small fraction of the Chinese population and the two data sources are unrelated. Our estimate for 2005 falls within the uncertainty limits reported by WHO and others, although our estimate of 4454 deaths is much lower than the 7800 deaths estimated by WHO, the United Nations Children’s Fund (UNICEF), the United Nations Population Fund (UNFPA) and The World Bank.31 The latter estimate is based on data from the sentinel surveillance system, where the reported MMR was accepted as the lower limit of uncertainty, twice the observed value was taken as the upper limit of uncertainty, and the midpoint of the uncertainty range was taken as the point estimate. In other words, estimates by WHO, UNICEF, UNFPA and The Word Bank inflated the reported numbers by 50% because of presumed underreporting.

There may be biases in both the denominator (live births) and the numerator (maternal deaths) of the MMR. Many studies have drawn attention to underreporting of births in national vital registration systems and fertility surveys,32,33 and the Routine Reporting System used in China is probably no exception. Underreported births tend to be girls or the outcome of pregnancies not approved by the family planning system. For this reason birth rates reported from censuses, routine reports or surveys are generally inflated upwards.3235 The yearly number of births counted through the Routine Reporting System was substantially lower than the number reported by the China Bureau of Statistics.36 This may be partly because the Routine Reporting System records births among women who are officially registered in their locality, so births among migrant women may be underreported. It is also possible that some home births were missed, although the high coverage of births in hospital and the extensive network of community health workers would limit this bias. For maternal deaths the situation is similar: underreporting may occur for deaths in early pregnancy or among migrant women, in areas where institutional birth rates are low, or because of an imperative for local officials to perform well and meet targets, although this is impossible to verify.

These potential biases are likely to affect the absolute levels of maternal mortality, but they do not necessarily affect time trends or differences between regions. We did not adjust the birth rates upward because the data for deaths and births were obtained from the same source, and arbitrarily inflating the number of births would have falsely lowered the MMR. The underreporting of births may not vary much with time or by region. However, if maternal deaths are more likely to be missed where institutional birth rates are low, the actual gap between richer and poorer provinces may be greater than what we found, particularly in the earlier years of our study period, when institutional births rates were lower.

Success stories in maternal health are greatly needed, and the positive lessons learned from China should give other countries the confidence that progress is possible. The factors that explain this success are multiple, but there is little doubt that a policy of institutional birth within a well-functioning health system has been a major contributor. The main challenge for China’s maternal health services is to maintain the gains already made for the majority, while prioritizing resources to the poorest areas. ■


We thank Dr Pang Ruyan for helpful comments.

Funding: This study was supported by the Maternal and Child Health Department, Ministry of Health, China. Professor Ronsmans was supported by a Research Fellowships under the 2007/08 United Kingdom/China Fellowships for Excellence programme.

Competing interests: None declared.