Geographic Variations and Temporal Trends in Cesarean Delivery Rates in China, 2008-2014

Key Points

Question  How has the rate of cesarean delivery in China changed between 2008 and 2014?

Findings  Cesarean rates increased from 28.8% in 2008 to 34.9% in 2014. However, there was substantial variation across China’s 31 provinces, with rates ranging from 4.0% to 62.5% in 2014.

Meaning  In 2014, the overall cesarean rate of 34.9% in China was close to the US rate of 32.4%, but there was large geographic variation in rates.

Importance  The increasing use of cesarean delivery is an emerging global health issue. Prior estimates of China’s cesarean rate have been based on surveys with limited geographic coverage.

Objective  To provide updated information about cesarean rates and geographic variation in cesarean use in China.

Design, Setting, and Data Sources  Descriptive study, covering every county (n = 2865) in mainland China’s 31 provinces, using county-level aggregated information on the number of live births, cesarean deliveries, maternal deaths, and perinatal deaths, collected by the Office for National Maternal & Child Health Statistics of China, from 2008 through 2014.

Exposures  Live births.

Main Outcomes and Measures  Annual rate of cesarean deliveries.

Results  Over the study period, there were 100 873 051 live births, of which 32 947 229 (32.7%) were by cesarean delivery. In 2008, there were 13 160 634 live births, of which 3 788 029 (28.8%) were by cesarean delivery and in 2014 there were 15 123 276 live births, of which 5 280 124 (34.9%) were by cesarean delivery. Rates varied markedly by province, from 4.0% to 62.5% in 2014. Despite the overall increase, by 2014 rates of cesarean delieries in 14 of the nation’s 17 “super cities” had declined by 4.1 to 17.5 percentage points from their earlier peak values (median, 11.4; interquartile range, 6.3-15.4). In 4 super cities with the largest decreases, there was no increase in maternal or perinatal mortality.

Conclusions and Relevance  Between 2008 and 2014, the overall annual rate of cesarean deliveries increased in China, reaching 34.9%. There was major geographic variation in rates and trends over time, with rates declining in some of the largest urban areas.

Overuse of cesarean delivery can jeopardize maternal and child health.^1,2 Concerns about “alarming”³ rates in China increased after the World Health Organization (WHO)⁴ reported that 46.2% of births were delivered by cesarean in the years 2007 and 2008, based on an analysis of 14 541 deliveries in 21 hospitals in 3 provinces.

With a population of approximately 1.37 billion in 2014, China’s health care use is of substantial global interest. Over the past 2 decades, the nation has invested in improving maternal and child health. The Law on Maternal and Infant Health Care was passed in 1995 to provide a high-quality hospital delivery to all women.⁵ Health insurance expansion began in 2003, to provide better financial access to maternity care.⁶Quiz Ref ID In 2009, a comprehensive health care reform plan included subsidies for in-hospital delivery to pregnant women in rural areas.⁷ In 2008, 89.9% of deliveries occurred in hospitals, and 87.9% of families had health insurance.⁶ By 2014, 99.6% of deliveries occurred in hospitals.⁸

China’s increasing cesarean rate has been spurred by many factors, including expanded access to hospital care, medicolegal concerns, financial incentives for physicians to perform cesareans, and parents’ focus on having a “perfect baby” under the one-child policy.^9,10 Since 2002, reducing the cesarean rate has been a national priority,¹¹ and a variety of policies, programs, and activities have emerged at the central government,¹² provincial or municipal,¹³ city,¹⁴ and city district or county levels.¹⁵ There is no national system of birth certificates in China, so national estimates of cesarean rates have been based on surveys with limited geographic coverage.^4,6,9 Given the marked diversity of geography, economy, and life circumstances throughout China, survey estimates are likely to be sensitive to the area sampled.

The Office for National Maternal & Child Health Statistics began to collect county-level national data on cesarean rates in 2008. The objectives of this study were to determine the overall rate and change in rate of cesarean deliveries, examine geographic variation, identify decreases in rates and in areas where decreases occurred, and assess changes in maternal and perinatal mortality.

The National Maternal & Child Health Statistics (NMCHS) data set, our primary source, has collected maternal and child health records since the mid-1990s. The Peking University Health Science Center Institutional Review Board did not need to approve this study because we used previously collected data containing no information on individual patients.

The dataset consists of county-level indicators of maternal and child health and health care service use, covering all counties and all births, primarily reported by county maternal and child health care hospitals. Indicators in the dataset serve as an important data source for government decision making. The total number of live births has long been considered gold standard information and has been used as basic information for the deployment of health resources nationally.^8,16 The set of indicators is reviewed every 2 to 3 years to reflect current health priorities. In 2008, information on cesarean deliveries was added in response to reports of increasing cesarean rates.

The total number of births in a county is reported in the database separately for resident and nonresident births. Resident births in a county refer to those to mothers or fathers with “Hukou” status (residential birth status) in the province in which that county is located. Nonresident births refer to those whose parents’ Hukou status is in other provinces. Typically, these are births to Chinese nationals working outside of their residential birth province. Although all births are reported, delivery mode (vaginal vs cesarean) is reported only for births to residents.

Data for in-hospital and out-of-hospital births are collected through China’s 3-tier system, the national maternal and child health data reporting network.¹⁷ Every county, every township, and every village countrywide has at least 1 person assigned to the network. In rural areas, village-level staff jointly work with township staff to collect birth information (including method of delivery) in the catchment area. Township staff report to the county every 1 to 6 months. In urban areas, the subdistrict physicians or clinicians collect individual-level data and transmit reports to the city district office every 1 to 6 months. County-level data are aggregated and reported annually. Until 2014, data were tabulated on standardized electronic spreadsheets at county offices and then reported to city and provincial offices and then to the Office for the National Maternal & Child Health Statistics by email. Since 2014, the county-level aggregated data has been reported directly via an online reporting system.

The Office for the National Maternal & Child Health Statistics oversees data quality, receives funds to train data collectors and managers, and provides annual training to provincial staff, visiting 2 to 3 provinces for quality control every year. Provincial staff train city staff, who train county staff. Provincial staff visit 2 to 3 cities, and city staff visit 2 to 3 counties for quality control every year. Quality control focuses primarily on the source data of all the indicators, particularly the number of live births and mortality data. Training on reporting the number of cesarean deliveries is provided to health workers, with a particular emphasis on counting the number of cesarean deliveries as the number of newborns in twin or multiple pregnancies (ie, the number of cesarean deliveries for twin births is 2). The reports of cesarean rates have not been validated by audit.

Secondary data sources included the China County Statistics online database¹⁸ and data on geographic boundaries in China.¹⁹

County-level information on the number of live births, cesarean deliveries, maternal deaths, stillbirths, and early neonatal deaths was extracted for all counties from 2008 through 2014. In all, 20 042 county-year records were extracted, with 24 records excluded for missing data. Prior to any analysis 2 authors familiar with the data identified obvious data entry errors, excluding 206 additional county-years (1.0%). Linkage with county gross domestic product (GDP) per capita data resulted in the removal of 63 additional county-years. In all, 19 749 county-years from 2865 counties (98.5% of county-years) were included in the analyses.

Cesarean rates were defined as the percentage of cesarean deliveries, maternal mortality as the number of maternal deaths (while pregnant or within 42 days of termination of pregnancy) per 100 000 live births, and perinatal mortality as the number of stillbirths (≥28 weeks of gestation) and early neonatal deaths (0-6 days) per 1000 births. As noted above, information about delivery mode is only available for resident women. Therefore, we report the total numbers of births and cesarean rates for resident women as the main outcome but also performed tests to assess the sensitivity of our findings to the inclusion of nonresident women, using varying assumptions about their delivery patterns.

Although data were reported at the county level, some analyses rates were calculated at higher levels as aggregations of counties.

We validated NMCHS county-level cesarean rates against rates reported in 2 sample-based studies.^4,6 We approximated the aggregate rates that we would have gotten from the NMCHS data, had we used each of the 2 sampling strategies, using published information about those strategies. For the 2009-2011 China National Health Services Survey, across the 94 sampled counties for the 2010 midpoint year, we calculated the arithmetic mean of the NMCHS cesarean rate.⁶ For the 2008 WHO survey, to each of the 21 sampled hospitals we assigned the NMCHS cesarean rate for the hospital’s corresponding geographic unit (city, county, or city district), and took the arithmetic mean of those 21 cesarean rates.⁴ We also validated the rates against audited rates from a major city (Tianjin in 2009 and 2011).²⁰

For each year, the national cesarean rate was calculated by aggregating county-level rates, weighting each county-level rate by the number of births to resident women in the county for that year. In other words, the national rate was

Σⁿ_i=1(w_i×CSrate_i)/Σⁿ_i=1w_i× 100%

for which n equals the number of counties; w_i, the number of births in the i^th county; and CSrate_i, the cesarean rate in the i^th county.

Quiz Ref IDWe compared secular trends in cesarean rates between subgroups of counties defined by (1) degree of urbanization (rural counties, general city counties, and super city counties); rural county refers to rural areas of a city, general city county refers to urban areas of a city with a population of less than 5 million, and super city county refers to urban areas of a city with a population of 5 million or more²¹; (2) cesarean rate in 2008, the start of the period (<30.0%, 30.0%-44.9%, and ≥ 45.0%); and (3) GDP per capita tertile in 2008. We performed stratified Poisson regression with generalized estimating equations, using the cesarean rate as the dependent variable and calendar year as the independent variable to examine trends in cesarean rates for subgroups. Significant differences in secular trends between subgroups were assessed by adding an interaction term between calendar year and corresponding characteristics defining subgroups. We used a geographic information system to display cesarean rates in 2008 and 2014 for all counties, and used funnel plots to display the extent to which the county-level variation in cesarean rates exceeded that of random fluctuations alone.²²

To assess the clustering effects in county-level cesarean rates within cities and provinces, we calculated intraclass correlation coefficients for county cesarean rates by using multilevel random intercept models, for which cesarean rates for counties nested within cities and provinces were considered as repeated measurements at level 1, each city represented a unit at level 2, and each province represented a unit at level 3.

We graphed rates of cesarean and perinatal and maternal mortality to visually inspect whether reductions in cesarean rates appeared to be accompanied by an increase in maternal and perinatal deaths in super cities, where high cesarean rates have been the historical norm.²³ We used Poisson regression with generalized estimating equations, with mortality as the dependent variable and calendar year as the independent variable, to examine whether perinatal or maternal mortality increased over the study period, in super cities with substantial reductions in rates over time (>15% from the peak rate), in which the greatest effect of substitution would be expected.

Statistical analyses were conducted using SPSS 20.0, using 2-tailed tests and a significance level of P < .05. Geographic analysis used ArcGIS 10.2 for Desktop Advanced, Student Edition.

We projected hypothetical overall (resident and nonresident) national rates under 2 assumptions: (1) rates for nonresident women are the same as for local resident women, and (2) rates for nonresident women are one-third lower than rates for local resident women, consistent with a recent report from the same period from Shanghai,²⁴ a super city.

In 2010, the arithmetic mean NMCHS rate across the 94 counties that participated in the 2009-2011 Health Services Survey was 37.3%, close to the national rate of 36.3% reported by the National Health Services Survey⁶ (eTable 1 in the Supplement). In 2008, the arithmetic mean NMCHS rate across the cities, counties, or city districts housing the 21 hospitals sampled by the WHO was 46.5%, close to the 46.2% rate reported by the WHO⁴ (eTable 2 in the Supplement). When cesareans were audited by the Tianjin health department, rates were 68.9% in 2009 and 56.7% in 2011.²⁰ Corresponding rates from the NMCHS were 71.5% and 60.5%, respectively.

From 2008 through 2014, there were 100 873 051 live births in the 2865 counties. Of these, 32 947 229 were by cesarean delivery, resulting in an overall national rate of 32.7%. Cesarean deliveries increased from 3 788 029 of 13 160 634 live births (a rate of 28.8%) in 2008 to 5 280 124 of 15 123 276 live births (a rate of 34.9%) in 2014, a mean increase of 1.0 (SD, 0.5) percentage point per year (Table 1). Growth was initially faster (1.8 percentage point increase from 2008 to 2009), but slowed by the end of the period (0.3 percentage point increase from 2013 to 2014). The range in provincial rates was large: Tibet had the lowest at 4% and Jilin the highest at 62.5% in 2014 (eTable 3 in the Supplement). There was substantial geographic clustering of rates (Figure 1). The intraclass correlation coefficient for county cesarean rates within cities and provinces was 0.752 in 2008 and 0.774 in 2014. In other words, 75.2% and 77.4% of the variation in county-level rates was accounted for by the regional clustering of cesarean rates within cities and provinces in 2008 and 2014, respectively. As shown by funnel plots (eFigure in the Supplement), variations in county cesarean rates far exceeded those expected by random fluctuation.

The high variability in rates is further quantified in Table 2. For example, in 2014, 19.6% of births were in counties where the cesarean rate was 20% or less, and 35.9% were in counties with rates greater than 40%. Figure 2 shows some county-level factors that were associated with levels and trends. As supported by the interaction test, there were diverse trends in cesarean rates among subgroups of counties defined by degree of urbanization, cesarean rate in 2008, and county GDP per capita tertile in 2008; corresponding P values for interaction tests were all <.001. Rates were highest in super cities and lowest in rural areas. Over the period, rates increased linearly in rural and general urban areas (P values for trend <.001) but declined beyond baseline rates in super cities (P for trend <.001) (Figure 2, A). Consistent with this, overall rates decreased over time in areas that had the highest baseline rates (P for trend <.001), but increased substantially in those with the lowest and intermediate rates (P for trend <.001) (Figure 2, B). Local economic conditions were associated with rates: high GDP per capita counties started with the highest rates, but showed the smallest increase over time (Figure 2, C).

In 2008 cesarean rates in the 17 super cities were between 18.2% and 68.8% (median, 48.7%; interquartile range [IQR], 40.6% to 64.1%); by 2014 rates in 4 super cities had declined by more than 15 percentage points from early in the study period (ie, 2008 or 2009); rates in 5 super cities decreased by 10.1 to 15 percentage points and in another 5 super cities by 4.1 to 10 percentage points from their peak values, which generally appeared in the middle of the study period (eTable 4 in the Supplement).

In the 4 super cities with the most pronounced reduction in cesarean rates, neither maternal nor perinatal mortality appeared to increase over the period (Figure 3). Statistical tests were consistent with this: the estimated annual percent change in maternal mortality was −12.8 (95% CI, −25.2 to 1.8; P = .08) in Beijing, −5.0 (95% CI, −18.5 to 10.8; P = .52) in Tianjin, −6.5 (95% CI, −20.5 to 10; P = .42) in Shanghai, and −5.5 (95% CI, −17.7 to 8.5; P = .42) in Hangzhou. For perinatal mortality, the corresponding estimates were −2.6 (95% CI, −4.9 to −0.4; P = .02) in Beijing, −7.1 (95% CI, −10.4 to −3.7; P < .001) in Tianjin, −2.2 (95% CI, −6.0 to 1.7; P = .26) in Shanghai, and −0.3 (95% CI, −4.4 to 4.0; P = .89) in Hangzhou.

In addition to the 100 873 051 births to resident women included in the main analysis, there were an additional 10 083 398 births to nonresident women (10% additional births). The hypothetical overall (resident and nonresident) national rates under the assumption that cesarean rates for nonresident women were the same as local resident women was 33.7% for the entire study period (eTable 5 in the Supplement). Under the assumption that cesarean rates were one-third lower for nonresident women, the corresponding rate was 32.4% (eTable 6 in the Supplement). Both are close to the rate in the resident population (32.7%). Year-specific rates under both assumptions are also close to those of the resident population (maximum difference, 1.3 percentage points), suggesting that births to nonresidents have a limited influence on the national rate. However, since nonresidents cluster in super cities, those rates are most sensitive to differential delivery patterns. For example, following the second assumption, the overall rate in Beijing dropped from 52.4% to 44.6% and in Shanghai from 60.9% to 50.8%.

Quiz Ref IDIn this data source with national coverage including 90% of all deliveries over 7 years in China, estimates of cesarean rates were substantially lower than those reported previously.⁴ In 2014, the rate was 34.9%, a figure more comparable with that of the United States (32.2% in 2014²⁵) than rates reported for such middle-income countries as Brazil (56%²⁶). Despite the overall increased trend, the growth rate had slowed by the end of study period.

Quiz Ref IDRates varied markedly across China. Even in 2014 they remained low in some rural areas, relative to the 19% rate that was recently estimated as a threshold target for improving maternal and child health.²⁷ Clinical data that would allow evaluation of the appropriateness of rates were not available, so definitive conclusions about overuse and underuse cannot be reached. But in some areas rates so far exceeded the 19% threshold that overuse seems likely, especially in urban areas, super cities, and areas with a high GDP per capita.

Rates decreased in 14 of the 17 super cities over the study period, with more than 15% decreases in 4 super cities. Four super cities with the greatest reductions did not show increases in rates of maternal or perinatal mortality, which is consistent with—but does not prove—that the substitution of cesarean with vaginal delivery was not associated with an increase in mortality. It is possible that concurrent improvements in health or health care outcomes for women and children masked negative health effects of substitution. Also, other adverse outcomes potentially relating to mode of delivery, such as neonatal injury, neonatal mortality, and maternal morbidity, were not measured, nor were trends in outcomes explored in areas where cesarean rates increased.

Over the study period, there were many national and local efforts to reduce the cesarean rate across China. Interventions included patient and clinician education, hands-on training for obstetricians and midwives, setting local target rates, and more recently financial or practice disincentives for failure to reach target goals.^13-15 Causal links to these policies and programs cannot be evaluated with these data for many reasons: information about implementation was sparse and clinical data were limited. However, reduction of cesarean rates in some areas may have coincided with some of these policies and programs.

During the study period, cesarean rates in poorer and rural counties increased. The percent of counties with rates of less than 10% was 18.5% in 2008, decreasing to 7.4% in 2014 (Table 2). This likely reflects the improvement in cesarean service capacity of health facilities as well as lower financial barriers to care in rural areas.^6,7

The analysis was based on aggregated data; no patient-level clinical data were available with which to adjust the rate estimates, and so some of the variability reported may reflect cross-area clinical heterogeneity in indications for cesarean. A recent report from the United Kingdom found that cesarean rate variability declined after clinical adjustment, but variance in patient population could not explain persistent wide variation in rates.²⁸Quiz Ref ID It seems unlikely that the extreme variability in China was largely accounted for by variations in clinical factors. Variability could also reflect a number of different health system factors, including differences in clinician training, facility capacity, or financial incentives. It could also reflect variations in local patient preferences. Examination of these are beyond the scope of the study.

This study used a census, rather than a sample. The NMCHS data emanate from a data collection system that has been in place since the 1990s, with quality controls at many levels. However, NMCHS cesarean rates have not yet been validated via internal audit. It was therefore reassuring that the NMCHS rates were consistent with rates derived from prior survey-based studies, after survey sampling strategies were taken into account. The NMCHS cesarean rates did not include nonresident women, who contribute approximately 10% of the births nationally. Evidence points to a lower rate of cesareans among nonresident women,^24,29,30 so we did not present analyses assuming higher rates for nonresidents. Given the known pattern of lower rates for nonresidents, it seems plausible that the rates reported herein are overestimates rather than underestimates of the true national rate.

Between 2008 and 2014, the overall annual rate of cesarean deliveries increased in China, reaching 34.9%. There was major geographic variation in rates and trends over time, with rates declining in some of the largest urban areas.

Corresponding Author: Jian-Meng Liu, MD, PhD, Institute of Reproductive and Child Health, Ministry of Health Key Laboratory of Reproductive Health, Peking University Health Science Center, No. 38, Xueyuan Rd, Haidian District, Beijing 100191, China (liujm@pku.edu.cn).

Author Contributions: Dr Liu had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs H-T Li and Luo are joint first authors.

Study concept and design: Trasande, Hellerstein, Liu, Blustein.

Acquisition, analysis, or interpretation of data: H-T Li, Luo, Hellerstein, Kang, J-X Li, Zhang, Liu.

Drafting of the manuscript: H-T Li, Luo, Liu, Blustein.

Critical revision of the manuscript for important intellectual content: All the authors.

Obtained funding: H-T Li, Liu.

Administrative, technical, or material support: Trasande, Kang, J-X Li, Zhang, Liu.

Study supervision: Trasande, Liu, Blustein.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Funding/Support: This study was funded by grants from the National Health and Family Planning Commission of China, 81273163 and 81571517 from the National Natural Science Foundation of China, and from the Beijing Young Talent Program.

Role of the Funder/Sponsor: The sponsors had no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank health workers who have contributed to the NMCHS data collection in provincial, city, county, township, and village levels, and acknowledge the support from You Li, MD, MPH, Institute of Reproductive and Child Health, Peking University Health Science Center, for his help in the preparation of the manuscript. The health workers and You Li received no compensation from this study.