Preterm Birth in China Between 2015 and 2016

Abstract

Objectives. To describe the incidence, risk factors, and potential causes of preterm birth (PTB) in China between 2015 and 2016.

Methods. The China Labor and Delivery Survey was a population-based multicenter study conducted from 2015 to 2016. We assigned each birth a weight based on the sampling frame. We calculated the incidence of PTB and the multivariable logistic regression, and we used 2-step cluster analysis to examine the relationships between PTB and maternal, fetal, and placental conditions.

Results. The weighted nationwide incidence of PTB was 7.3% of all births and 6.7% of live births at 24 or more weeks of gestation. Of the PTBs, 70.5% were born after 34 weeks and 42.7% were iatrogenic. Nearly two thirds of all preterm births were attributable to maternal, fetal, or placental conditions, and one third had unknown etiology.

Conclusions. This study provided information on the incidence of PTB in China and identified several factors associated with PTB. The high frequency of iatrogenic PTB calls for a careful assessment and prudent management of such pregnancies, as PTB has short- and long-term health consequences.

Preterm birth (PTB), defined as a delivery before 37 completed weeks of gestation, is one of the most common causes of perinatal mortality.^1,2 In 2015, PTB-related complications were the leading causes of child under-5 mortality and the second most common cause of child death after congenital abnormalities in China.^3,4 PTB also has long-term consequences such as increased risks of cerebral palsy, impaired learning, visual disorders, hearing loss, and chronic noncommunicable diseases later in life.^5,6 In addition, PTB and its associated sequelae have enormous negative impacts on the family and society.^1,7

According to the World Health Organization (WHO) Report in 2019, 14.8 million babies were born too soon with the global average PTB rate of 10.6%, and China had more than 1.1 million PTBs with the rate of 6.9%.⁸ However, there is heterogeneity in the PTB rate reported in different studies. In 2005, a large study showed that, among 42 139 live births, the incidence of PTB was 7.8% from 77 hospitals in 16 provinces in China.⁹ Zou et al. also conducted a multicenter hospital-based survey including 107 905 live births in 39 hospitals from 14 provinces, and the incidence of PTB was 7.1% in 2011.¹⁰ Yet, these figures differed substantially from an incidence of 11.0% of singleton live births from 63 tertiary hospitals participating in the National Neonatal Network in 23 provinces in China between 2011 and 2014.¹¹ More studies with a larger number of hospitals and wider regional distribution are warranted. Furthermore, a detailed analysis of the epidemiology of PTB was often lacking. Using the data from the China Labor and Delivery Survey, a multicenter study that covered most areas of China, we aimed to describe the epidemiology of PTB at the national level.

METHODS

The China Labor and Delivery Survey was a multicenter cross-sectional survey throughout the country between March 1, 2015, and December 31, 2016. Participation was solicited through obstetric conferences and networks. Hospitals that expressed an interest were asked to provide basic information about the hospitals. Only those with 1000 or more deliveries per year were eligible. Six weeks within a 12-month period for hospitals with an annual delivery of at least 6000, or 10 weeks for hospitals with an annual delivery fewer than 6000 per year were randomly selected for data collection. Within each selected week, all births delivered at 24 or more completed weeks of gestation or weighing 500 grams or more at birth were eligible. Medical records were retrieved and reviewed, and information was extracted by trained research nurses. This methodology has been previously used in the WHO Global Survey of Maternal and Perinatal Health and the WHO Multi-Country Survey of Maternal and Newborn Health.^12,13

A data coordination center was established, which was responsible for establishment, management, and maintenance of the database and Web site, coordination among hospitals, investigators’ training, and quality control of the data. A logic check function was programmed during the establishment of the database for preventing missing items and data entry errors. Data administrators checked the data regularly and contacted the hospital investigators to fill in the omissions or correct the errors.

A total of 89 hospitals in 25 (out of 34) provinces and autonomous regions in China were involved in the current analysis. We defined PTB as any birth before 37 completed weeks of gestation determined by the best available obstetric estimate, which mostly relied on an early pregnancy ultrasound in combination with the last menstrual period.¹⁴ The main outcome was the incidence of PTB calculated as the percentage of any birth (live or still) between 24 weeks 0 days and 36 weeks 6 days of gestation. On the basis of the etiology, we divided PTB into spontaneous PTB and iatrogenic (medically induced) PTB. We further separated spontaneous PTB as spontaneous onset of labor and preterm premature rupture of the fetal membranes; iatrogenic PTB referred to the induction of labor or elective caesarian birth before 37 completed weeks of gestation for maternal or fetal indications or other nonmedical reason.^15,16 On the basis of the gestational age, we grouped PTB into 4 categories: extremely early PTB (24–27 weeks), early PTB (28–31 weeks), moderate PTB (32–33 weeks), and late PTB (34–36 completed weeks).^1,16

To better represent the epidemiology of PTB in China overall, we calculated a weight for each birth in the survey by using the number of deliveries in each province from the 2016 China Statistical Yearbook, compiled by the National Bureau of Statistics of China (http://www.stats.gov.cn/tjsj/ndsj/2016/indexch.htm). We considered hospital level as a poststratification factor, and we stratified the annual number of childbirths in each city by hospital levels. We assigned each birth a weight based on the inverse probability weighting taking into account the number of births in the province with the same hospital level and the number of records reviewed in the hospital with the same hospital level.¹⁷ We used the bootstrap method to calculate 95% confidence intervals (CIs) of the incidence of PTB.¹⁸

China is divided into 7 administrative regions: northeast, north, east, central, south, southwest, and northwest.¹⁰ We grouped maternal age into younger than 20, 20.0 to 24.9, 25.0 to 29.9, 30.0 to 34.9, and 35 years and older. We classified maternal prepregnancy body mass index (BMI; defined as weight in kilograms divided by the square of height in meters [kg/m²]) as underweight (< 18.5 kg/m²), normal weight (18.5–24.9 kg/m²), and overweight or obesity (≥ 25 kg/m²) according to the WHO BMI classification.¹⁹ We categorized maternal education level as low (illiterate, primary school, and junior school), middle (high school, technical school, and junior college), and high (college or higher degree).

Maternal medical conditions included maternal diseases (chronic hypertension, diabetes mellitus, gestational diabetes mellitus, renal and cardiac disease, immune system disease, thyroid disease, and asthma), hypertensive disorders in pregnancy (HDP; including gestational hypertension, preeclampsia/eclampsia, chronic hypertension superimposed by preeclampsia), sexually transmitted disease, history of miscarriage or stillbirth, history of PTB, and conception by assisted reproductive technology. Placental abnormalities included placenta abruption, placenta previa, and chorioamnionitis. We also examined fetal conditions such as fetal distress, multiple pregnancy, fetal anomaly, sex indetermination, and antepartum stillbirth.

We used multivariable logistic regression (nnet package, multinom function in R version 3.4.4, R Foundation for Statistical Computing, Vienna, Austria) to examine the relationships between PTB and maternal, fetal, and placental conditions. We also carried out a 2-step cluster analysis, a hierarchical agglomerative approach to combine homogenous cases into clusters and detect latent relationships between PTBs with more than 1 maternal, fetal, or placental condition.²⁰ We used SPSS version 22.0 (IBM, Somers, NY).

RESULTS

We initially included a total of 77 879 birth records in the analysis from 89 hospitals in 25 provinces. Of these, 22 hospitals were secondary and 67 were tertiary hospitals; 36 hospitals had an annual delivery of fewer than 5000 births and 13 hospitals more than 10 000 births. We further excluded birth records with gestational age less than 24 weeks or more than 44 weeks (n = 2155) and those with missing maternal age or older than 45 years (n = 134). A total of 75 590 births constituted the final sample, including 74 667 live births, 877 stillbirths, and 46 births with unknown outcomes.

The unweighted incidence of PTB was 9.5 per 100 all births (95% CI = 9.3%, 9.7%) and 8.7 per 100 live births (95% CI = 8.5%, 8.9%). After weighting adjustment, the weighted incidence of PTB was 7.3 per 100 all births (95% CI = 7.0%, 7.6%) or 6.7 per 100 live births (95% CI = 6.4%, 7.0%) in China for the period of 2015 to 2016.

Figure 1 and Figure A (available as a supplement to the online version of this article at http://www.ajph.org) show that there was a substantial difference in the incidence of PTB by geographic regions. At the provincial level, the weighted incidence of all PTB ranged from 5.1% (95% CI = 4.2%, 6.3%) in Liaoning Province to 16.7% (95% CI = 13.8%, 19.7%) in Shanxi Province. Central China had the lowest incidence (6.1%; 95% CI = 5.4%, 6.8%) whereas southwest China has the highest incidence (11.4%; 95% CI = 10.5%, 12.2%). For PTB in live births, northeast China had the lowest incidence (5.2%; 95% CI = 4.4%, 6.0%) and southwest China still had the highest incidence (10.5%; 95% CI = 9.7%, 11.3%). The geographic maps of PTB rates are presented in Figure B (available as a supplement to the online version of this article at http://www.ajph.org).

FIGURE 1—

Incidence of Preterm Birth (Weighted), by Province, for (a) All Births and (b) Live Births: China, 2015–2016

Figure 2 summarizes the subgroups of PTB. Iatrogenic PTB accounted for the largest proportion of PTB (42.7%), followed by spontaneous onset of labor (29.5%) and preterm premature rupture of the fetal membranes (27.8%). The distribution of the potential causes of PTB differed by the hospital level. Spontaneous onset of labor (38.0%) was the most common type of PTB in the secondary hospitals whereas iatrogenic PTB (47.7%) accounted for the most in the tertiary hospitals. Extreme preterm, early, moderate, and late PTBs accounted for 5.9%, 11.0%, 12.5%, and 70.5%, respectively. There appeared to be a higher portion of extremely early PTB (6.5%) and late PTB (75.7%) in the secondary hospitals than in the tertiary hospitals (5.5% and 66.4%, respectively). In primiparas, the weighted PTB rate was 7.4%, and spontaneous PTB accounted for 60.7% of all PTB. In multiparas, the corresponding rates were 7.3% and 53.6%, respectively.

FIGURE 2—

Percentage of Preterm Births (Weighted) by Hospital Level for (a) Etiology and (b) Gestational Age: China, 2015–2016

Note. pPROM = preterm premature rupture of membranes; PTB = preterm birth.

Table 1 presents unadjusted and adjusted odds ratios with 95% CIs for PTB. Younger or older maternal age, overweight or obesity, maternal diseases, HDP, history of miscarriage or stillbirth, history of PTB, placenta previa, placenta abruption, chorioamnionitis, fetal distress, multiple pregnancy, fetal anomaly, male fetus, and antepartum stillbirth were associated with an increased risk of PTB. A stratified analysis was conducted by spontaneous versus iatrogenic PTB and the results are provided in Table A (available as a supplement to the online version of this article at http://www.ajph.org). Sexually transmitted disease, previous miscarriage or stillbirth, previous PTB, and chorioamnionitis tended to be associated more with spontaneous PTB whereas HDP, placenta previa, multiple pregnancy, fetal anomaly, and antepartum stillbirth showed a stronger association with iatrogenic PTB.

TABLE 1—

Distribution of Characteristics and Their Associations With Preterm Birth: China, 2015–2016

	Unweighted		Weighted (n = 7 785 677)
Characteristics	Total, No.	Preterm Birth, No.	Proportion,a %	Incidence, %	OR (95% CI)	AOR (95% CI)
Hospital level
Secondary	17 435	1 036	59.6	5.4	1 (Ref)	1 (Ref)
Tertiary	58 155	6 143	40.4	10.0	1.94 (1.93, 1.95)	1.92 (1.90, 1.93)
Hospital type
Maternity	40 216	3 694	52.4	6.9	1 (Ref)	1 (Ref)
General	35 375	3 485	47.6	7.7	1.12 (1.12, 1.13)	0.96 (0.95, 0.97)
Maternal age, y
< 20	1 083	167	2.1	9.9	1.73 (1.70, 1.75)	1.73 (1.68, 1.77)
20–24	11 109	1 027	18.9	7.1	1.20 (1.19, 1.21)	1.05 (1.04, 1.06)
25–29	35 060	2 845	46.0	6.0	1 (Ref)	1 (Ref)
30–34	19 848	2 049	22.6	8.5	1.47 (1.46, 1.48)	1.13 (1.12, 1.14)
≥ 35	8 490	1 091	10.3	10.5	1.85 (1.84, 1.87)	1.35 (1.34, 1.37)
Prepregnancy BMI
Underweight (< 18.5 kg/m2)	8 154	858	9.4	7.5	1.24 (1.23, 1.25)	0.96 (0.95, 0.97)
Normal (18.5–24.9 kg/m2)	41 062	3 379	51.8	6.1	1 (Ref)	1 (Ref)
Overweight or obesity (≥ 25 kg/m2)	7 014	590	8.5	6.0	0.98 (0.97, 0.99)	1.04 (1.03, 1.05)
Mother’s level of educationb
Low	16 663	2 026	32.3	7.7	1.04 (1.03, 1.04)	1.07 (1.06, 1.08)
Middle	30 129	2 779	38.0	7.4	1 (Ref)	1 (Ref)
High	20 730	1 612	17.3	6.8	0.91 (0.91, 0.92)	0.84 (0.83, 0.84)
Maternal diseasec
No	61 344	5 526	82.1	6.8	1 (Ref)	1 (Ref)
Yes	14 246	1 653	17.9	9.6	1.45 (1.44, 1.46)	1.20 (1.18, 1.21)
Chronic hypertension	352	96	0.3	24.4	4.12 (3.99, 4.25)	2.49 (2.37, 2.61)
Diabetes mellitus	764	129	0.8	13.0	1.93 (1.88, 1.97)	1.13 (1.09, 1.17)
Gestational diabetes	7 764	973	8.2	10.9	1.65 (1.64, 1.66)	1.23 (1.21, 1.24)
Thyroid disease	1 466	144	1.2	10.1	1.44 (1.41, 1.47)	1.12 (1.09, 1.16)
Heart disease	296	50	0.3	15.6	2.35 (2.26, 2.44)	1.62 (1.53, 1.71)
Renal disease	134	33	0.2	16.4	2.52 (2.40, 2.64)	1.22 (1.10, 1.34)
Hypertensive disorders in pregnancy
No	71 274	6 237	95.7	6.8	1 (Ref)	1 (Ref)
Yes	3 299	855	3.9	20.1	3.46 (3.43, 3.49)	3.05 (3.01, 3.09)
Sexually transmitted diseases
No	5 062	7 106	99.0	7.2	1 (Ref)	1 (Ref)
Yes	427	51	0.8	14.7	2.21 (2.16, 2.26)	1.33 (1.28, 1.38)
Previous miscarriage or stillbirth
No	68 923	6 315	91.7	7.0	1 (Ref)	1 (Ref)
Yes	6 666	864	8.3	10.8	1.61 (1.60, 1.62)	1.16 (1.15, 1.18)
Previous preterm birth
No	74 515	6 916	98.6	7.1	1 (Ref)	1 (Ref)
Yes	1 074	263	1.4	21.5	3.57 (3.52, 3.63)	2.76 (2.70, 2.83)
Assisted reproductive technology
No	70 084	6 188	93.8	6.9	1 (Ref)	1 (Ref)
Yes	5 213	960	6.0	13.3	2.07 (2.05, 2.09)	1.04 (1.03, 1.05)
Placenta abruption
No	74 934	6 961	99.4	7.2	1 (Ref)	1 (Ref)
Yes	512	199	0.5	35.0	6.96 (6.82, 7.11)	5.56 (5.39, 5.74)
Placenta previa
No	74 512	6 751	99.0	7.0	1 (Ref)	1 (Ref)
Yes	1 022	418	0.9	38.5	8.29 (8.16, 8.41)	6.54 (6.38, 6.70)
Chorioamnionitis
No	75 356	7 134	99.7	7.3	1 (Ref)	1 (Ref)
Yes	234	45	0.3	16.2	2.46 (2.37, 2.55)	3.02 (2.87, 3.17)
Fetal distress
No	72 376	6 718	95.3	7.0	1 (Ref)	1 (Ref)
Yes	3 033	409	4.4	11.8	1.76 (1.74, 1.78)	1.35 (1.33, 1.37)
Multiple pregnancy
No	71 361	5 372	95.5	6.1	1 (Ref)	1 (Ref)
Yes	3 980	1 774	4.2	33.4	7.67 (7.61, 7.73)	8.20 (8.10, 8.30)
Fetal anomaly
No	74 152	6 695	97.8	7.0	1 (Ref)	1 (Ref)
Yes	545	278	0.6	40.4	9.02 (8.85, 9.19)	3.14 (3.04, 3.24)
Sex
Female	34 794	3 020	45.8	6.6	1 (Ref)	1 (Ref)
Male	39 902	3 986	52.7	7.8	1.20 (1.20, 1.21)	1.29 (1.28, 1.30)
Antepartum stillbirth
No	73 990	6 390	97.6	6.7	1 (Ref)	1 (Ref)
Yes	877	685	0.9	70.1	32.74 (32.22, 33.28)	22.56 (22.03, 23.11)

Note. AOR = adjusted odds ratio (calculated by multivariable logistic analysis adjusted for all variables); BMI = body mass index; CI = confidence interval; OR = odds ratio.

^aPercentages may not sum to 100 because of missing values.

^bMother’s level of education: low (illiterate, primary school, and junior school), middle (high school, technical school, and junior college), and high (college or higher degree).

^cMaternal diseases: chronic hypertension, diabetes mellitus, gestational diabetes mellitus, renal and cardiac disease, immune system disease, thyroid disease, and asthma.

The 2-step cluster analysis identified 5 clusters of potential causes of PTB (Table 2). The value of the silhouette measures of cohesion and separation statistic was 0.5, and the clustering was considered to be fair. The largest cluster (cluster 1, comprising 34.6% of all PTBs) was idiopathic and was not associated with any underlying conditions considered. Cluster 2 (24.4% of all PTBs) was composed of a mixed group, dominated by placental and fetal conditions, such as fetal distress (30.2%), antepartum stillbirth (28.6%), placenta previa (19.6%), history of PTB (16.4%), fetal anomaly (13.7%), placenta abruption (7.8%), sexually transmitted disease (7.1%), and chorioamnionitis (2.5%). Cluster 3 (16.5% of PTBs) was also a mixed cluster manifested by history of miscarriage or stillbirth (56.1%) and HDP (53.6%). In cluster 4 (12.8% of PTBs), all cases had maternal diseases. Cluster 5 contained 11.6% of all PTBs with a mixture of conditions including multiple pregnancy (79.8%) and use of assisted reproductive technology (45.2%). Further details of all row and column percentages for each cluster are provided in Tables B through P and Figure C (available as supplements to the online version of this article at http://www.ajph.org).

TABLE 2—

Distribution of the 5 Clusters of Preterm Births According to Maternal, Fetal, or Placental Conditions (Weighted): China, 2015–2016

Cluster	No. (%)	Main Condition (%)
1	187 035 (34.6)	None
2	131 758 (24.4)	Fetal distress (30.2), antepartum stillbirth (28.6), placenta previa (19.6), history of preterm birth (16.4), fetal anomaly (13.7), placenta abruption (7.8), sexually transmitted disease (7.1), chorioamnionitis (2.5)
3	89 286 (16.5)	History of miscarriage or stillbirth (56.1), hypertensive disorders in pregnancy (53.6)
4	69 310 (12.8)	Maternal diseasesa (100.0)
5	62 928 (11.6)	Multiple pregnancy (79.8), assisted reproductive technology (45.2)
Combined	540 317 (100.0)

^aMaternal diseases: chronic hypertension, diabetes mellitus, gestational diabetes mellitus, renal and cardiac disease, immune system disease, thyroid disease, and asthma.

DISCUSSION

In our study, we found that the incidence of PTB (24–36 completed weeks of gestation) was 7.3 per 100 total births (95% CI = 7.0%, 7.6%) and 6.7 per 100 live births (95% CI = 6.4%, 7.0%) in China. Of all PTBs, 42.7% were iatrogenic. By the 2-step cluster analysis, we identified 5 clusters of potential causes of PTB. Of all PTBs 34.6% were idiopathic.

We also observed a great variation in the incidences of PTB among Chinese regions. Higher incidences were recorded in the western region (11.4% for southwest China and 10.3% for northwest China). This pattern is consistent with previous studies, in which the incidence was 10.3% in the southwest.¹⁰ These regions are less developed than the other parts of the country, supporting the evidence that low socioeconomic status has consistently been linked to PTB.²¹ Also, the western regions have higher latitude and more minority populations. How the interaction of genetics and environment contributes to PTB remains to be elucidated.²² By contrast, height has been inversely associated with the risk of PTB.²³ Because women in northeast China are generally taller,²⁴ this might partially explain why the northeast region had the lowest PTB rate.

By contrast, the incidence of PTB was significantly higher in the tertiary hospitals (10.0%) with more advanced facilities and specialists than in the secondary hospitals (5.4%). This is because the majority of high-risk women are transferred to the tertiary hospitals for better perinatal care. The high proportion of tertiary hospitals participating in our survey may explain the decrease in PTB rate from 9.5% (unweighted) to 7.3% (weighted). Given that the incidence of PTB varied greatly by geographic distribution and hospital level, a weighted estimate is likely to be more accurately representative in China.

PTB is widely considered to be a complex syndrome, consisting of several clusters of potential causes and being initiated by multiple mechanisms. Risk factors of PTB have already been well documented in the literature, and our results were consistent with previous studies.^15,16,25 We also identified 5 clusters of potential causes related to maternal, fetal, and placental conditions. Surprisingly, the largest cluster, accounting for one third of PTBs in our study, was not associated with any obvious maternal or fetal conditions. This finding is also consistent with the International Fetal and Newborn Growth Consortium for the 21st Century (INTERGROWTH-21st) Project, which included 5828 PTBs among 53 871 deliveries in 8 countries. It identified 12 clusters, and the most common cluster, accounting for 34.6% of the total PTB, was idiopathic PTB.²⁶ The high proportion of idiopathic PTB continues to challenge the conventional prevention strategies currently available.

On the contrary, iatrogenic PTB is induced artificially. It is the most common cause in China, accounting for 42.7% of PTBs in our study and 39.4% in the study by Zou et al.¹⁰ Globally, it is around 30% according to a review in 2008.¹⁵ This proportion has been shown to be increasing in the United States, from 26% in the 1990s to 37% in the mid-2000s, along with increasing overall PTB rates.^27,28 Although medical interventions are generally considered beneficial to reduce stillbirths and neonatal deaths in cases in which fetal or maternal compromise is evident,²⁷ overuse of interventions may have led to unnecessary iatrogenic PTBs. As even late preterm deliveries have been reported to be independently associated with adverse neonatal outcome,²⁹ clinicians and health care providers need to routinely assess high-risk pregnancies and ensure appropriateness of interventions that can result in provider-initiated PTB.^30,31 In addition to the prevention and management of these maternal, fetal, and placental conditions, defining exact circumstances in which the balance of benefits and risks favors delivery over expectant management should also be one of the top priorities for future research to reduce PTB rate in China.

The incidence of PTB varies widely in the literature,^9–11 partially attributable to the fact that studies used different methods to calculate the PTB rate. Although the upper limit of gestational age is clearly defined—36 weeks plus 6 days—the lower boundary could vary from 20 to 28 weeks in different countries.⁸ For example, an investigation in 19 European countries included only live births after 22 weeks in their analysis,³² whereas a Chinese study excluded births before 28 weeks’ gestation.¹⁰ As birth is legally defined as the expulsion of the product of conception at 28 weeks or later in China, PTB rates were likely to have been underestimated in previous studies.³³ Furthermore, the PTB rates vary widely among regions and hospitals in China. Unweighted estimates of PTB rates may have limited representativeness and comparability.

In addition, the incidence of PTB can be affected by the accuracy of gestational age estimation. It is well known that an estimate of gestational age based on the last menstrual period is subject to various errors and tends to overestimate the PTB rate, while estimation by ultrasound in early pregnancy offers a superb alternative.³⁴ Thus, in countries where last menstrual period is commonly used for gestational age estimation without correction by early ultrasound examination, the PTB rate is likely to be overestimated. All hospitals that participated in our survey provide ultrasound examination services, and we could speculate that the vast majority of gestational ages were estimated or confirmed by ultrasound in early pregnancy.

Moreover, there is currently no consensus as to whether stillbirth should be included in the calculation of the rate of PTB. For example, “Born Too Soon” took both singleton and multiple live births into consideration,¹ whereas a study conducted in 6 high-income countries in North America and Europe included only singleton live births delivered at 22 or more completed weeks of gestation.³⁰ As multiple pregnancy and stillbirth are common causes of PTB, the PTB rate without these births is likely to be underestimated in these countries. On the other hand, reliable stillbirth data are not always readily available, especially in developing countries. As a compromise, PTB may be counted only among live births. To make data comparable among countries and studies, a clear definition of PTB is essential in every study. We further suggest that in studies in which reliable stillbirth data are available, it will be useful to provide PTB rates both in all births and in live births for comparison and future meta-analysis.

Limitations and Strengths

Despite the large sample size, our study has some limitations. First, we have relatively little information on maternal environment and behavior, such as smoking, physical activity, and eating habits. Thus, we do not have a complete profile of risk factors for PTB.

Second, our participants were not a random sample of all births in China. Some provinces had no participating hospitals; we had more tertiary than secondary hospitals and no primary hospitals. Such a study sample may lead to bias in the estimation of PTB incidence to some degree. However, few primary hospitals deliver babies in China and the annual delivery volume in those hospitals is usually low.³⁵ There is neither a PTB registry system nor large-scale epidemiology survey in rural China. Our data suggest that PTB is negatively correlated with the degree of economic development, which is consistent with data from other countries. Thus, the exclusion of primary hospitals may have underestimated the rate. On the other hand, the incidence of PTB was significantly higher in tertiary hospitals with advanced facilities and subspecialists than in secondary hospitals because the majority of high-risk pregnant women are transferred to tertiary hospitals for better perinatal care. Thus, the referral system may have led to an overestimate of the PTB rate. To improve the representativeness of our study population, we stratified hospital level as the primary sampling units when computing the weighting value to reduce these biases.

Third, a higher incidence of PTB in rural and less developed regions of China may be in part attributable to late initiation of prenatal care in some women and shortage of ultrasound service in early pregnancy. Finally, the incidence of PTB is also determined by obstetric practice style. A same condition may be managed differently by different obstetricians, resulting in different likelihood of PTB. In addition, suboptimal medical record keeping in some facilities might have adversely affected data quality. However, we tried to minimize this impact by using uniform study methods and definitions of the variables collected across the facilities and by training data collectors.

Public Health Implications

PTB is a crucial global public health issue. Our survey demonstrated that the incidence of PTB remained high and varied widely across geographic regions in China, which may serve to inform policymaking and resource reallocation to address burdens of PTB. We call for careful assessment and prudent management of high-risk pregnancies, as PTBs have short- and long-term neonatal consequences. Meanwhile, strategies to avoid unnecessarily medical interventions need to be continually reinforced based on the very high rate of iatrogenic PTB. As the 2-child policy gathers momentum in China, the proportion of pregnant women with advanced age, comorbidities, and assisted reproductive technology increases further. We identified and clustered a number of risk factors and potential causes from our analysis. For better prevention and management of PTB, research on the potential mechanisms is still needed.