Preterm birth (< 37 weeks) is one of the strongest predictors of perinatal and infant mortality in the United States and other industrialized countries and confers an increased burden of serious morbidity and long-term neurocognitive disability in comparison with birth at term gestation.1 Although prevention of preterm birth remains a perinatal priority in industrialized countries,2–4 clinical and community-based interventions have fallen short of reducing preterm birth.3 In fact, the rate of preterm birth has increased in the United States5–6 and in several other countries7–8 in the recent decade.
Preterm birth is a multifactorial and heterogeneous outcome. A focus on overall trends may conceal important patterns within preterm birth subtypes, namely, preterm birth following ruptured membranes, medically indicated preterm birth, and spontaneous preterm birth.9 Intrauterine infection plays a major role in preterm birth associated with preterm rupture of membranes.3 Medically indicated preterm birth, on the other hand, is typically necessitated in the presence of fetal compromise and for maternal indications such as preeclampsia and other pregnancy complications.10–14 Preterm birth following spontaneous onset of labor is a significant contributor to all preterm birth, yet its etiology is poorly understood.9 Another important factor in preterm birth rates in the United States is the prevailing racial disparity, with blacks at an approximately 2-fold higher risk of delivering preterm than whites.5,15 Equally noteworthy is the race-specific temporal trend in preterm birth rates; the rate has declined among blacks but has increased among whites recently.5,12
We examined temporal trends in preterm birth in the recent decade among singleton births in the United States, focusing separately on each of the 3 preterm birth subtypes and maternal race. We hypothesized that the rate of medically indicated preterm birth would have increased in the recent decade in the United States and that this increase would be associated with a decline in perinatal mortality. We restricted our study focus to singleton births because preterm birth rates and changes in rates differ drastically between singleton and multiple births.5,7
MATERIALS AND METHODS
We used data from the United States singleton natality and fetal death files, 1989 through 2000, and the infant death data files from 1989 through 1991 and 1995 through 2000, to examine temporal trends in preterm birth subtypes and perinatal mortality. Information in these files corresponds to data abstracted from birth certificates of liveborn infants and from fetal and infant death certificates and was assembled by the National Center for Health Statistics of the Centers for Disease Control and Prevention.16 Infant death data files for 1992 through 1994 were not linked to the corresponding live births by the National Center for Health Statistics, and appropriate data were not available for analysis.
Gestational age data, reported in completed weeks, were predominantly based on the date of the last menstrual period, calculated as the difference between the date of the menstrual period and the date of delivery. In a small fraction of births (< 5%) that did not contain a valid date for the menstrual period, or when the menstrual-based gestational age was incompatible with birth weight, a clinical estimate of gestation, also contained on vital records, was used instead.17 If a valid month and year of the last menstrual period was available but the day was missing, gestational age (also in completed weeks) was statistically imputed by the National Center for Health Statistics before release of the data.18 The replacement of menstrual-based gestational age by a clinical estimate, as well as imputation of missing gestational age, was performed consistently for all years (1989 through 2000).
Trends in preterm birth (< 37 weeks) were assessed among singletons after excluding fetuses that weighed less than 500 g or those that were delivered before 22 weeks of gestation. Preterm birth was classified into 3 distinct subtypes: preterm birth following ruptured membranes; medically indicated preterm birth, defined as a preterm birth that followed iatrogenic intervention (labor induction or a primary or repeat cesarean delivery); and spontaneous preterm birth, defined as preterm birth that was neither associated with ruptured membranes nor was medically indicated.
Four check box items on the live birth and fetal and infant death certificates were used to assign preterm birth subtypes.17 Preterm birth subtypes was assigned based on the following hierarchy: preterm birth following ruptured membranes, medically indicated preterm birth, and, finally, spontaneous preterm birth. The item “ruptured membranes” accompanied by delivery at less than 37 weeks was used to identify women who delivered preterm following ruptured membranes. The items “primary cesarean,” “repeat cesarean,” or “labor induction,” accompanied by delivery at less than 37 weeks were combined to identify women who delivered preterm following medical indications. The item corresponding to labor induction represents an attempt to induce labor, regardless of its success.
We first examined absolute and relative trends in preterm birth and their subtypes between 1989 and 2000. This was done for all births (all races), as well as for births to whites and blacks. Absolute trends in preterm birth subtypes were based on calculating the change in rates between the 2 periods (1989 and 2000). Relative changes in preterm birth (and their subtypes) between 1989 and 2000 were quantified by using relative risk (RR) with 95% confidence interval (CI).
We examined the contribution of changes in preterm birth subtypes to perinatal mortality by using an ecological design. Perinatal mortality19 included stillbirths at 22 weeks or later plus deaths among liveborn infants within the first 28 days (ie, neonatal deaths). This analysis was done by using logistic regression with perinatal death as the dependent variable. The data for this analysis was first stratified by state within the United States (50 levels) and period (2 levels, 1989 and 2000), which resulted in 100 strata. For each stratum, we then determined the preterm birth proportion (ie, preterm birth rate by subtype for each state in 1989 and in 2000), as well as the proportions of women aged less than 20 years and 35 years or older, the proportion of primigravid women, the proportion of single women, and the proportion of women with fewer than 12 years of completed schooling. We regressed the preterm birth variable (in addition to state, the period variable, and confounders) against perinatal death. From these models, we obtained the RR and 95% CI, denoting the change in perinatal mortality rates between 1989 and 2000, before and after adjustment for temporal changes in preterm birth subtypes. This analysis was repeated for stillbirths and neonatal deaths.
The ecological analysis was necessary to avoid confounding by indication, which occurs in nonexperimental studies that estimate the effect of a clinical intervention20 using individual-level analyses. Such confounding is likely because labor induction and cesarean delivery are typically used in pregnancies with fetal compromise and because antepartum stillbirths are often delivered by inducing labor.
In addition to preterm birth at less than 37 weeks, we replicated the entire analysis for preterm birth at less than 35 weeks. Furthermore, the impact of preterm birth at less than 37 and at less than 35 weeks (and the subtypes) on perinatal mortality, stillbirth, and neonatal death were examined after restricting the analysis to nonmalformed births. Statistical tests were 2-tailed, and variability of the effect measure (ie, RR) was assessed by the width of its 95% CI.
Of all singleton births in the United States between 1989 and 2000, we sequentially excluded 324,745 births with birth weight less than 500 g, 314,555 with gestational age less than 22 completed weeks (to exclude possible temporal bias in the registration of livebirth and stillbirth at the borderline of viability21), 513,401 births with missing gestational age, and 194,462 births that ended at 46 weeks or later (owing to the high proportion of errors in the menstrual estimate of gestational age at these extreme gestational ages22). After all exclusions, 46,375,578 women who delivered singleton births between 1989 and 2000 in the United States remained for analysis.
The study was approved by the Ethics Review Committee of the Institutional Review Board of the UMDNJ–Robert Wood Johnson Medical School, New Brunswick, New Jersey. All statistical analysis was carried out with SAS 8.2 (SAS Institute, Cary, NC) on the UNIX platform.
Preterm birth (< 37 weeks) among singletons increased from 10.0% in 1989 to 10.4% in 2000 (Fig. 1A,). Preterm birth resulting from ruptured membranes declined from 1.3% in 1989 to 0.9% in 2000, medically indicated preterm birth increased from 2.6% to 3.8%, and spontaneous preterm birth decreased from 6.1% in 1989 to 5.7% in 2000. The yearly changes in preterm birth subtypes relative to the rate in 1989 are shown in Figure 1B.
Among whites, the overall preterm birth rate increased from 8.3% in 1989 to 9.4% in 2000, whereas among blacks, the rate decreased from 18.5% to 16.2% between the same periods (Table 1). The increase in the overall rate of preterm birth among whites was largely the result of a 55% increase (RR 1.55) in medically indicated preterm birth, whereas the decline in the rate of preterm birth among blacks was due to a 37% relative decrease (from 2.3% in 1989 to 1.5% in 2000) in preterm birth following ruptured membranes and a 27% decrease (from 12.1% in 1989 to 9.1% in 2000) in spontaneous preterm birth. Medically indicated preterm birth increased by 32% (from 4.1% in 1889 to 5.6% in 2000) among blacks, although the increase was much larger among whites than among blacks (55% versus 32%). When the analyses were restricted to births without congenital anomalies, the patterns of preterm birth trends were similar to those seen among all births.
Among preterm births to white women in 2000, the proportions attributable to preterm birth following ruptured membranes, medical indications, and spontaneous preterm birth were 8%, 38%, and 54%, respectively. The corresponding proportions among blacks were 3%, 35%, and 62%, respectively.
Trends in overall stillbirth rates, in stillbirths within the preterm birth range, and in stillbirth within each subtype of preterm birth are shown in Figure 2. Overall perinatal mortality rates and perinatal mortality rates within each preterm birth subtype are shown in Table 2. Among whites, overall perinatal mortality rates declined from 7.8 per 1,000 total births in 1989 to 5.4 per 1,000 total births in 2000, a relative decline of 30% (RR 0.70). Adjustment for confounders did not alter this decline appreciably (Table 3). When the period change in perinatal mortality was further adjusted for preterm birth following ruptured membranes or for spontaneous preterm births, there was a 2% change in mortality rates (ie, the 29% decline in perinatal mortality was reduced to 27%). However, when the perinatal mortality rate was adjusted for medically indicated preterm birth, the period change declined to 21% (RR 0.79). In other words, if the rate of medically indicated preterm birth had not increased by 55% among whites between 1989 and 2000 (Table 1), the perinatal mortality rate would have declined only by 21% (and not by the observed 29%). Thus, the largest decline in perinatal mortality among whites was due to increases in medically indicated preterm birth.
Among blacks, the observed decline in perinatal mortality between 1989 (14.8 per 1,000 total births) and 2000 (10.8 per 1,000 total births) was 27% (RR 0.73). Changes in the distribution of confounders did not affect this trend substantially (Table 2). When this trend was further adjusted for preterm birth following ruptured membranes or spontaneous preterm birth, the temporal decline in the perinatal mortality rate was reduced to 16% (RR 0.84) and 17% (RR 0.83), respectively. When the mortality change was adjusted for medically indicated preterm birth, the period decline was reduced to 20% (RR 0.80). The largest effect on perinatal mortality among blacks, therefore, occurred secondary to reductions in preterm birth after ruptured membranes and reductions in spontaneous preterm birth. Analyses excluding births affected by congenital anomalies yielded similar patterns.
Supplementary analyses carried out after defining preterm birth as that occurring before 35 weeks of gestation showed patterns similar to the overall and race-specific trends in preterm birth subtypes and in their impact on perinatal mortality. Similarly, analyses of stillbirth and neonatal mortality revealed patterns similar to those seen for overall perinatal mortality.
We have shown that the temporal increase in preterm birth among whites in the United States was largely driven by an increase in medically indicated preterm birth, whereas among blacks, the preterm birth rate declined because of a decrease in spontaneous preterm birth and preterm birth following ruptured membranes. The most striking observation in our study pertains to the association between changes in preterm birth subtypes and trends in perinatal mortality. The decline in perinatal mortality between 1989 and 2000 among whites was most strongly associated with increases in medically indicated preterm birth, whereas among blacks the mortality decline was most closely associated with reductions in preterm birth following ruptured membranes and spontaneous preterm birth. The relatively larger increase in medically indicated preterm birth among whites compared with blacks is also noteworthy. These racial differences in obstetric intervention rates have to be evaluated against a background of differences in sociodemographic and other risk factors, including access to prenatal care. The increase in preterm birth rates in most industrialized countries has been viewed as disappointing and deemed a “failure” of efforts to reduce preterm birth.4,23 We noted a rise in medically indicated preterm birth—an observation that corroborates findings from other countries.7,12,24,25 This increase in preterm birth has been associated with declines in rates of stillbirth7 and perinatal mortality.26 Our study demonstrates that this association is primarily a consequence of increases in medically indicated preterm birth.
Obstetric intervention (labor induction and cesarean delivery) has increased among singleton11 and twin13,14,26 pregnancies, especially at later gestational ages, namely at 34 weeks or later, and to a lesser extent at 32 weeks or later.12–14,26 With dramatic advances in neonatal care, these interventions are being increasingly used at preterm gestation to reduce perinatal mortality (and serious morbidity) in cases where fetal, and less commonly, maternal compromise is evident. The increase in medically indicated preterm birth should therefore be viewed as being beneficial, insofar as such intervention reduces stillbirths and neonatal deaths, as demonstrated in our study. In fact, given contemporary trends in maternal characteristics, including increases in maternal age27,28 and obesity,29,30 one can expect this trend toward a higher rate of medically indicated preterm birth to continue. The fact that medically indicated preterm birth rates did not rise among blacks to the extent they did among whites warrants attention. Nevertheless, the decline in preterm birth among blacks due to declines in spontaneous preterm birth and preterm birth following ruptured membranes is encouraging.
The different subtypes of preterm birth share common determinants, and hence an increase in a particular subtype may result in a decrease in another subtype. For instance, an increased awareness of the signs and symptoms of labor can lead to an increase in spontaneous preterm birth at the expense of preterm birth following ruptured membranes. Similarly, women with early spontaneous preterm labor who are subsequently delivered by cesarean may have been identified as medically indicated preterm birth in our study. However, the frequency of such misclassification is likely to be low and unlikely to impact our findings because we focused on temporal changes and not absolute rates of each preterm birth subtype. The differences, and potential overlap, between preterm birth subtypes also mean that etiologic studies and community programs aimed at preventing preterm birth should examine preterm birth both in terms of its overall frequency and also by subtype to more fully recognize and understand potential changes. This is particularly true given the general failure of preterm birth prevention programs and also the recent identification of progesterone as a promising medical intervention for preventing recurrent spontaneous preterm birth.31,32
The limitations of our study include those typical of studies that use information from large databases. Errors in menstrual estimate of gestational age are likely to have affected our results to a small extent33,34 if the magnitude of these errors remained constant over the study period. In addition, temporal trends in preterm birth at the population-level have been affected by changes in the assessment of gestational age from assessments based largely on menstrual period to those based on early ultrasonography.8 In fact, there is some evidence to suggest that gestational age estimation may have differentially improved among blacks compared with whites, especially at 28–31 weeks of gestation.35 Because most preterm birth occurs at 32–36 weeks of gestation, this change is unlikely to have substantially affected the results of our study. In our study, imputation of missing gestational age and replacement of menstrual estimate of gestational age by a clinical estimate, although consistently performed by the National Center for Health Statistics for all years, may have affected trends in the preterm birth rates.18 A sensitivity analysis examining the effect of these changes (carried out through a reanalysis after excluding the imputed gestational ages) revealed essentially similar patterns in preterm birth trends. Some of the clinical conditions (such as premature rupture of membranes and labor induction) noted on vital statistics may be subject to misclassification. Studies validating the accuracy of obstetric procedures on birth certificates against hospital medical records have demonstrated a moderate-to-high sensitivity in the reporting of operative vaginal deliveries from birth certificate data.36,37
It is not possible to draw definitive causal inferences from our study, most notably in relation to the effect of medically indicated preterm birth on perinatal mortality. The data files used in our study do not distinguish between antepartum and intrapartum fetal deaths. Nevertheless, the beneficial effect of medically indicated preterm birth in preventing perinatal death, which was observed in our study, offers reasonable evidence that such obstetric intervention is saving compromised fetuses.
The large population-based nature of the study and the focus on relative, rather than absolute, changes reduces the impact of the above mentioned limitations. Confounding by indication20 is the primary reason for preferring an ecological, as opposed to an individual-level, logistic regression analysis. The large size of the study enabled us to detect important differences in race-specific trends in preterm birth subtypes and perinatal mortality.
The increase in overall preterm birth in the United States is associated with a concurrent increase in medically indicated preterm birth, and such intervention is associated with declines in perinatal mortality. Because prevention of preterm birth remains one of the top goals under the nation’s Healthy People 2010 objective,38 efforts need to be targeted toward women who spontaneously deliver preterm. Similarly, the persistently large, and disappointing, black-white disparity in preterm birth in the United States and associated perinatal mortality merits attention.
1. Goldenberg RL, Hoffman HJ, Cliver SP. Neurodevelopmental outcome of small-for-gestational age infants. Eur J Clin Nutr 1998;52(suppl):S54–8.
2. Slattery MM, Morrison JJ. Preterm delivery. Lancet 2002;360:1489–97.
3. Goldenberg RL, Jobe AH. Prospects for research in reproductive health and birth outcomes. JAMA 2001;285:633–9.
4. Buekens P, Klebanoff M. Preterm birth research: from disillusion to the search for new mechanisms. Paediatr Perinat Epidemiol 2001;15(suppl 2):159–61.
5. Demissie K, Rhoads GG, Ananth CV, Alexander GR, Kramer MS, Kogan M, et al. Trends in preterm birth and neonatal mortality among blacks and whites in the United States from 1989 to 1997. Am J Epidemiol 2001;154:307–15.
6. Branum AM, Schoendorf KC. Changing patterns of low birthweight and preterm birth in the United States, 1981–98. Paediatr Perinat Epidemiol 2002;16:8–15.
7. Joseph KS, Kramer MS, Marcoux S, Ohlsson A, Wen SW, Allen A, et al. Determinants of preterm birth rates in Canada from 1981 through 1983 and from 1992 through 1994. N Engl J Med 1998;339:1434–9.
8. Kramer MS, Platt R, Yang H, Joseph KS, Wen SW, Morin L, et al. Secular trends in preterm birth: a hospital-based cohort study. JAMA 1998;280:1849–54.
9. Savitz DA, Blackmore CA, Thorp JM. Epidemiologic characteristics of preterm delivery: etiologic heterogeneity. Am J Obstet Gynecol 1991;164:467–71.
10. Yawn BP, Wollan P, McKeon K, Field CS. Temporal changes in rates and reasons for medical induction of term labor, 1980–1996. Am J Obstet Gynecol 2001;184:611–9.
11. ZhangJ, Yancey MK, Henderson CE. U.S. national trends in labor induction, 1989–1998. J Reprod Med 2002;47:120–4.
12. Joseph KS, Demissie K, Kramer MS. Obstetric intervention, stillbirth, and preterm birth. Semin Perinatol 2002;26:250–9.
13. Ananth CV, Joseph KS, Kinzler WL. The influence of obstetric intervention on trends in twin stillbirths, United States 1989–99. J Matern Fetal Neonatal Med 2004;15:380–7.
14. Ananth CV, Joseph KS, Smulian JC. Trends in twin neonatal mortality in the United States: influence of obstetric intervention. Am J Obstet Gynecol 2004;190:1313–21.
15. Berkowitz GS, Blackmore-Prince C, Lapinski RL, Savitz DA. Risk factors for preterm birth subtypes. Epidemiology 1998;9:279–85.
16. National Center for Health Statistics. 1989–2000 linked birth/infant death data set. NCHS CD-ROM series 20, No 3–20. Hyattsville (MD): United States Department of Health and Human Services, Centers for Disease Control and Prevention.
17. Taffel SM, Ventura SJ, Gay GA. Revised U.S. certificate of birth—new opportunities for research on birth outcome. Birth 1989;16:188–93.
18. Taffel S, Johnson D, Heuse R. A method of imputing length of gestation on birth certificates. Vital Health Stat 2 1982;93:1–11.
19. Cunningham GF, Gant NF, Leveno KJ, Gilstrap LC, Hauth JC, Wenstrom KD. Williams obstetrics. 21st ed. New York (NY): McGraw-Hill; 2001.
20. Wen SW, Kramer MS. Uses of ecologic studies in the assessment of intended treatment effects. J Clin Epidemiol 1999;52:7–12.
21. Kramer MS, Platt RW, Yang H, Haglund B, Cnattingius S, Bergsjo P. Registration artifacts in international comparisons of infant mortality. Paediatr Perinat Epidemiol 2002;16:16–22.
22. Kramer MS, McLean FH, Boyd ME, Usher RH. The validity of gestational age estimation by menstrual dating in term, preterm, and postterm gestations. JAMA 1988;260:3306–8.
23. Kramer MS. Preventing preterm birth: are we making any progress? Yale J Biol Med 1997;70:227–32.
24. Papiernik E, Zeitlin J, Rivera L, Bucourt M, Topuz B. Preterm birth in a French population: the importance of births by medical decision. BJOG 2003;110:430–2.
25. Foix-L’Helias L, Blondel B. Changes in risk factors of preterm delivery in France between 1981 and 1995. Paediatr Perinat Epidemiol 2000;14:314–23.
26. Joseph KS, Allen AC, Dodds L, Vincer MJ, Armson BA. Causes and consequences of recent increases in preterm birth among twins. Obstet Gynecol 2001;98:57–64.
27. Martin JA, Hamilton BE, Ventura SJ, Menacker F, Park MM. Births: final data for 2000. Natl Vital Stat Rep 2002;50:1–101.
29. Joseph KS, Young DC, Dodds L, O’Connell CM, Allen VM, Chandra S, et al. Changes in maternal characteristics and obstetric practice and recent increases in primary cesarean delivery. Obstet Gynecol 2003;102:791–800.
30. Cnattingius S, Lambe M. Trends in smoking and overweight during pregnancy: prevalence, risks of pregnancy complications, and adverse pregnancy outcomes. Semin Perinatol 2002;26:286–95.
31. Meis PJ, Klebanoff M, Thom E, Dombrowski MP, Sibai B, Moawad AH, et al, for the National Institute of Child Health and Human Development Maternal–Fetal Medicine Units Network. Prevention of recurrent preterm delivery by 17 alpha-hydroxyprogesterone caproate [published erratum appears in N Engl J Med. 2003;349:1299]. N Engl J Med 2003;348:2379–85.
32. da Fonseca EB, Bittar RE, Carvalho MH, Zugaib M. Prophylactic administration of progesterone by vaginal suppository to reduce the incidence of spontaneous preterm birth in women at increased risk: a randomized placebo-controlled double-blind study. Am J Obstet Gynecol 2003;188:419–24.
33. Yang H, Kramer MS, Platt RW, Blondel B, Breart G, Morin I, et al. How does early ultrasound scan estimation of gestational age lead to higher rates of preterm birth? Am J Obstet Gynecol 2002;186:433–7.
34. Gjessing HK, Skjaerven R, Wilcox AJ. Errors in gestational age: evidence of bleeding early in pregnancy. Am J Public Health 1999;89:213–8.
35. Vahratian A, Buekens P, Bennett TA, Meyer RE, Kogan MD, Yu SM. Preterm delivery rates in North Carolina: are they really declining among non-Hispanic African Americans? Am J Epidemiol 2004;159:59–63.
36. Buescher PA, Taylor KP, Davis MH, Bowling JM. The quality of the new birth certificate data: a validation study in North Carolina. Am J Public Health 1993;83:1163–5.
37. Parrish KM, Holt VL, Connell FA, Williams B, LoGerfo JP. Variations in the accuracy of obstetric procedures and diagnoses on birth records in Washington State, 1989. Am J Epidemiol 1993;138:119–27.
38. Healthy People 2010. 2nd ed. Washington, DC: U.S. Department of Health and Human Services; 2000.