The increase in cesarean deliveries has caused concern in many Western countries. In the United States, for example, the Public Health Service has set national objectives to reduce the use of cesarean delivery. 1,2 Although the US rate declined from 1989 to 1996, it has increased sharply since 1996, and reached a new high of 24.4% in 2001. 3 A similar increase in the cesarean delivery rate has been observed in the United Kingdom, where the current rate is 23%, and the National Institute for Clinical Excellence is developing guidelines to limit the use of the procedure. 4 Although governments are trying to curb the use of cesarean delivery, some data suggest that lowering of the cesarean delivery rate may increase the occurrence of adverse maternal and infant outcomes, especially uterine rupture and birth injury. 5 No consensus has been reached on what the right number of cesarean deliveries should be because little information is known about the difference in birth outcomes associated with higher or lower cesarean delivery rates. 4–7
The appropriate design of outcome studies that can address this issue has been problematic. Comparing outcomes between vaginal and cesarean deliveries is inherently biased because patients who deliver by the two methods have different risks. To avoid this problem, we take advantage of the difference in practice styles among community obstetricians to explore the effect of frequent or restrained use of cesarean delivery on maternal and infant outcomes. Previous studies have shown that there is wide variation in the cesarean delivery rate among obstetricians and that physician's style of practice is a major determinant of the decision to perform the procedure, independent of maternal and infant risk factors. 8–13 Risk differences between women who happen to fall into high cesarean delivery and low cesarean delivery practices are small, and they can be taken into account because the a priori risk differences that are relevant (maternal race, parity, previous cesarean delivery, hospital, etc) are mostly recorded in the birth certificate and hospital discharge databases. Therefore, here we group obstetricians by their cesarean delivery rates and compare risk-adjusted birth outcomes achieved by the physician groups. Grouping obstetricians provides the large number of births needed to study rare outcomes and avoids the undue influence of any one obstetrician. Record linkage of hospital discharge data, birth or fetal death certificate data, and infant death certificate data provides extensive covariate information, which we use to adjust risk differences between the populations served.
MATERIALS AND METHODS
We studied singleton births (live births and fetal deaths) at New Jersey hospitals between January 1, 1996 and December 31, 1997. We linked infant vital certificate data to the corresponding maternal and infant hospital discharge records, using previously validated probabilistic record linkage methodology. 14 Data fields were cross-checked for linking quality. Records with low linking agreements were examined manually and duplicated records were deleted. The final linked database included 212,675 births, representing 97% of all certified singleton births.
Maternal–fetal medicine specialists were excluded before the grouping of physicians because they specialize in managing high-risk pregnancies that are related to both higher rates of cesarean delivery and more adverse perinatal outcomes (7849 deliveries). Physicians who attended fewer than 100 births during the 2-year period were also excluded to assure that physicians included in the study had adequate practice in obstetrics (20,019). In addition, births with the following characteristics were excluded: infants not delivered by obstetricians (8411) deliveries by mothers who were not New Jersey residents (3497), mothers who were transferred from another hospital to the delivery hospital (1051), and deliveries with gestational age less than 23 weeks or birth weight less than 500 g (553). After the exclusions, 171,295 singleton births and 791 physicians remained for analysis.
Physicians, identified by their state license number from the hospital discharge file, were divided into three groups according to their cesarean delivery rates during the study period: low (less than 18%), medium (18–27%), and high (greater than 27%). This classification was based on an average of 216 deliveries per physician. The cutoff rates for the physician groups were based on quartile distribution of physician cesarean delivery rates. Physicians whose rates fell in the top quartile were classified as having a high rate, and those in the bottom quartile as having a low rate. Perinatal mortality and a spectrum of other perinatal outcomes were compared among the groups of physicians. The study was approved by the institutional review boards of the New Jersey State Department of Health and Senior Services and the Robert Wood Johnson Medical School.
Perinatal mortality, birth injury, and uterine rupture are the outcomes of interest. Perinatal mortality included early neonatal mortality (infant death within 7 days of birth) and fetal mortality (included both intrapartum fetal deaths and fetal deaths at admission). Neonatal mortality was defined as infant death within the first 28 days of birth. Birth injuries were categorized into two major groups: intracranial hemorrhage (subdural or cerebral hemorrhage, intraventricular hemorrhage, and subarachnoid hemorrhage) and nerve injury (facial nerve injury and brachial plexus injury), based on the diagnosis codes on infant hospital discharge using Classification of Diseases, 9th Revision (ICD-9), Clinical Modification. The analysis of birth injuries was restricted to live births with gestational ages of 34 weeks or greater. Uterine ruptures were identified from the diagnosis codes on the maternal hospital discharge records.
Information on mode of delivery and maternal complications was obtained from maternal hospital discharge records because research has shown that hospital discharge data are more complete and accurate than birth certificate data. 15 Cesarean delivery was determined using the ICD-9, Clinical Modification procedure code 74, excluding 74.3 (removal of intraperitoneal embryo) and 74.91 (hysterotomy to terminate pregnancy). A cesarean delivery was classified as repeat if a diagnosis of previous cesarean delivery (ICD-9, Clinical Modification code 654.2) was present. Vaginal births after cesarean delivery (VBACs) were defined as from those women who had a previous cesarean delivery but delivered vaginally this time. Trial of labor in women with prior cesarean deliveries was identified as occurring with those who delivered vaginally or those who had cesarean delivery but had diagnosis or procedure codes indicating labor in the hospital discharge record. Hospitals were classified according to state-designated level of obstetric care: basic, intermediate, intensive, and perinatal center.
Multivariable logistic regression models were used to assess the relationship between physician groups and the outcomes variables. The medium-rate group was used as the reference because it contained the majority of physicians and, hence, represented current prevailing obstetric practice. Potential confounding variables included in the final models were selected based on both a priori knowledge and statistical significance. 16 Known confounders that have been identified by previous research, such as maternal age, race, and parity, were included in the model regardless of statistical significance. Imbalances in patient characteristics that were clinically relevant to the perinatal outcomes were also included. Other variables such as maternal education and insurance status were selected to the model using the forward selection method with an entry criterion of P < .20. Interactions between physician groups and the confounding variables were assessed by the likelihood ratio test at the significance level of .05. Adjusted relative risks (RRs) and their 95% confidence intervals (CIs) were calculated. All analyses were performed using the SAS System (SAS Institute Inc., Cary, NC).
One hundred sixty-eight physicians and 36,063 births were classified to the low-rate group, 404 physicians and 90,069 births to the medium-rate group, and 219 physicians and 45,163 births to the high-rate group. The medium- and the high-rate groups were comparable in most of the maternal and infant demographic variables and medical risks (Table 1). Relative to these two groups, mothers in the low-rate group were more likely to be younger (under 20), to be black, to have Medicaid as primary insurance, to smoke or drink, and to begin prenatal care late or have no care at all. They were less likely to complete high school or to be married. The three groups had similar proportions of nulliparous women.
Breech pregnancies, preeclampsia or eclampsia, placenta previa, previous cesarean delivery, fetal distress, and dystocia were recorded more frequently in physician groups performing more cesarean deliveries. There were no significant differences in percentage of pregnancies complicated by diabetes, placental abruption, herpes, or cord prolapse. The physician groups were similar in the distribution of births by birth weight and gestational age categories. Frequencies of serious congenital anomalies also did not differ among the three groups. Physicians in the high-rate group were less likely to deliver at perinatal centers.
The overall cesarean delivery rate in the study sample was 23.3%. The rate of primary cesarean delivery was 16.3%, the rate of repeat cesarean delivery 67.5%, and the rate of VBAC 32.5%. Cesarean delivery rates varied substantially among the physician groups, being 14.2% in the low-rate, 22.5% in the medium-rate, and 32.2% in the high-rate group. Groups with higher total cesarean delivery rates performed more cesarean deliveries in women with and without previous cesarean deliveries (primary cesarean delivery rates of 10.1%, 15.9%, and 22.5% for the low-, medium-, and high-rate groups, respectively; repeat cesarean delivery rates of 48.7%, 65.3%, and 80.5% for the three groups, respectively). The rate of trial of labor in women with a previous cesarean delivery was 80% higher in the low-rate group than in the high-rate group (66.6% versus 37.7%). Low-rate physicians had higher rates of VBAC: 51.3% in the low-, 34.7% in the medium-, and 19.5% in the high-rate group. Lower rate physicians used more vacuum deliveries (7.1% versus 6.5% in the medium- and 5.1% in the high-rate groups). The medium-rate group used more forceps (3.76%) than the low- (2.34%) and high-rate (2.54%) groups.
High-rate physicians performed more cesarean deliveries in all types of patients, regardless of maternal and infant characteristics (Table 2). Rates of cesarean delivery were about twice as high for high-rate physicians than for low-rate physicians in women with previous cesarean deliveries, fetal distress, dystocia, diabetes, preeclampsia or eclampsia, and herpes. However, cesarean delivery rates for breech, placenta previa, and cord prolapse were more similar for the three physician groups. Variations in cesarean delivery rates were less pronounced among infants younger than 33 weeks' gestation.
The three physician groups were comparable in perinatal mortality before and after adjustment of patient risks (Figure 1). Low and very low birth weight infants delivered by physicians in the high-rate group did not have a lower risk of mortality (Table 3). On the contrary, there was an unexpected finding of excess fetal death in the high-rate group for low birth weight infants, even after adjusting for maternal and fetal risk factors. This increase in fetal mortality contributed to the increase in perinatal deaths for the high-rate group in this birth weight category. Similar findings were also observed in an analysis stratified by gestational age categories. There were 34 intrapartum fetal deaths in the study sample: ten (0.03%) in the low-, 16 (0.02%) in the medium-, and eight (0.02%) in the high-rate group.
In addition, the low- and medium-rate physician groups did not differ in the rate of neonatal intensive care unit (NICU) admission (3.7% for both groups). The high-rate group had a lower rate of NICU admission than the medium group (3.0% versus 3.7%), but it was not statistically significant after adjusting for patient risk factors.
There were 156 (nine per 10,000 births) intracranial hemorrhages and 379 (22 per 10,000 births) nerve injuries among live births with gestational ages of 34 weeks or greater. Infants delivered by high-rate physicians had the fewest intracranial hemorrhages (Figure 2A). Risk of intracranial hemorrhage was significantly higher (adjusted RR 1.53; 95% CI 1.07, 2.19) for infants delivered by low-rate physicians than for those delivered by medium-rate physicians. The medium- and high-rate groups had similar rates of intracranial injury. Average lengths of hospital stay were 11.2 days for infants with intracranial hemorrhage and 3 days for those with nerve injuries. Nerve injuries did not differ between the low- and medium-rate groups (Figure 2B). Relative to the medium-rate group, infants delivered by high-rate physicians had 26% less risk of nerve injury (adjusted RR 0.74; 95% CI 0.57, 0.96).
Uterine rupture occurred in 143 (eight per 10,000 deliveries) women in the study sample. Among these, 130 (91%) women labored. For women with a previous cesarean delivery, uterine rupture rate was higher in women who attempted labor (74 per 10,000 births) than in those with elective repeat cesarean deliveries, (three per 10,000 births). Relative to deliveries by medium-rate physicians, deliveries by low-rate physicians had decreased overall risk of uterine rupture (adjusted RR 0.56; 95% CI 0.34, 0.92; Figure 3). This was partly because fewer women had a prior cesarean delivery in the low-rate group (Table 1). Occurrences of uterine rupture were similar in the high- and medium-rate groups (Figure 3). Among women with a prior cesarean delivery, the rate of uterine rupture did not differ significantly (3.7, 4.6, and 2.9 per 1000 births for the low-, medium-, and high-rate groups, respectively). Among women without a prior cesarean delivery, the rates of uterine ruptures were 1.6, 4.0, and 4.0 per 10,000 births for the low-, medium-, and high-rate groups, respectively. There were five perinatal deaths associated with uterine rupture, with one in the low-, two in the medium-, and two in the high-rate group. Among women who had uterine rupture, two had hysterectomy in the low-, five in the medium-, and six in the high-rate group. There were no maternal deaths in women who had a uterine rupture.
In a population-based analysis, we found that nonsub-specialist obstetricians with low, medium, and high cesarean delivery rates achieved comparable perinatal mortality, and that higher cesarean delivery rates for the delivery of low or very low birth weight infants did not improve their survival. However, infants delivered by low-rate physicians were at increased risk of intracranial injury. The rate of uterine rupture was not increased in patients attended by physicians with lower cesarean delivery rates.
Although several previous studies have also attempted to examine whether physicians who perform more cesarean deliveries have better birth outcomes, 8–13 small sample sizes and the rarity of major adverse outcomes have limited statistical power and, hence, the interpretation of results. Moreover, patient risk factors that potentially influence the probability of cesarean delivery and adverse birth outcomes were not adequately controlled.
Our results on perinatal mortality confirm suggestions from international comparisons and smaller previous studies that a higher cesarean delivery rate is not necessarily associated with better perinatal outcomes 17,18 and that the cesarean delivery rate could be lowered without increase in mortality. 19–22 As in previous research, we found that lower cesarean delivery rates were mainly achieved by performing fewer cesarean deliveries for dystocia and by attempting more VBACs. The increased use of cesarean delivery for dystocia was the most important contributor to the rise in the cesarean delivery rate during the 1970s and 1980s. 23 Our finding of lack of benefit of high cesarean delivery rate in perinatal mortality for very low birth weight infants (adjusted RR 0.79; 95% CI 0.54, 1.15) contrasts with the only other study on perinatal mortality by birth weight categories, where maternity units with higher cesarean delivery rates had lower perinatal mortality for infants in this birth weight category. 24 Neither study found that higher cesarean delivery rates resulted in reduced perinatal mortality for low birth weight and normal birth weight infants. The finding of increased fetal death in the high-rate group for low birth weight infants may not be related to physician cesarean delivery policy because 93% of all fetal deaths were coded as occurring before labor. It could be a chance finding or might be related to differences in other aspects of management of pregnancy such as fetal monitoring.
The finding of increased risk of intracranial hemorrhage for infants delivered by physicians with lower cesarean delivery rates in this study is supported by results from a hospital-based study 11 and a population-based analysis on the occurrence of intracranial injury by modes of delivery. 25 The latter found that instrumental deliveries and cesarean delivery during labor were associated with increased risk of intracranial injury relative to spontaneous vaginal delivery, whereas the rate of injury from cesarean delivery before labor was not higher. The study concluded that a long, dysfunctional labor, rather than the mode of delivery, was the major risk factor for intracranial injury. Similarly, in our study the increase of intracranial injury in the low-rate group cannot be explained by greater number of instrumental deliveries because the rates of instrumental delivery were similar in the low- and medium-rate groups (9.7% versus 10.6%) for infants included in the analysis of birth injury. However, more women labored in the low-rate group, and low-rate physicians probably allowed more prolonged labor, as reflected by a lower cesarean delivery rate for dystocia. This is consistent with the idea that long labor contributes to the risk of intracranial injury.
The rate of uterine rupture in our study was comparable to that from statewide analyses in California and Washington based on hospital discharge records. 26,27 Consistent with other reports, we found that the risk of rupture was higher in women having a trial of labor after a previous cesarean delivery than in women having an elective repeat cesarean delivery. 26–28 We also found that women attended by low-rate physicians were less likely to have uterine rupture. This was partly explained by the lower number of women with a previous cesarean delivery in the low-rate physician group (Table 1). There was no excess perinatal mortality associated with reported uterine rupture in the low cesarean delivery rate group, suggesting that coding for this complication was not biased.
Our study is based on administrative data that are known to have limitations. However, certification of perinatal deaths is essentially complete, and the coding of cesarean delivery in computerized hospital discharge summaries is considered to be accurate in reflecting medical records. 29
These results suggest that policies organized to lower cesarean delivery rates should focus on appropriate use of cesarean delivery rather than just targeting of lower rates. Although lower cesarean delivery rates do not appear to be associated with increased perinatal mortality, death is not the only important outcome. The possibility that lower cesarean delivery rates may increase the risk for birth injury, especially intracranial hemorrhage, deserves attention.
1. Healthy People 2000, National health promotion and disease prevention objectives. DHHS publication no. (PHS) 91-50212. Washington: U.S. Department of Health and Human Services, Public Health Services, 1991.
2. Healthy People 2010. Conference ed. Vol I–II. Washington: U.S. Department of Health and Human Services, 2000.
3. Martin JA, Park MM, Sutton PD. Births: Preliminary data for 2001. Natl Vital Stat Rep 2002;50(10):1–20.
4. Feinmann J. How to limit cesareans on demand—too NICE to push? Lancet 2002;359:774.
5. Sachs BP, Kobelin C, Castro MA, Frigoletto F. The risk of lowering the cesarean-delivery rate. N Engl J Med 1999; 340:54–7.
6. What is the right number of cesarean sections [editorial]? Lancet 1997;349:815.
7. Treffers PE, Pel M. The rising trend for cesarean section. BMJ 1993;307:1017–8.
8. Goyert GL, Bottoms SF, Treadwell MC, Nehra PC. The physician factor in cesarean birth rates. N Engl J Med 1989;320:706–9.
9. DeMott RK, Sandmire HF. The Green Bay cesarean section study, I: The physician factor as a determinant of cesarean birth rates. Am J Obstet Gynecol 1990;162:1593–9.
10. Berkowitz GS, Fiarman GS, Mojica MA, Bauman J, de Regt RH. Effect of physician characteristics on the cesarean birth rate. Am J Obstet Gynecol 1989;161:146–9.
11. Guillemette J, Fraser WD. Differences between obstetricians in cesarean section rates and the management of labor. Br J Obstet Gynaecol 1992;99:105–8.
12. Lagrew DC, Adashek JA. Lowering the cesarean section rate in a private hospital: Comparing of individual physicians’ rates; risk factors, and outcomes. Am J Obstet Gynecol 1998;178:1207–14.
13. Poma PA. Effects of obstetrician characteristics on cesarean delivery rates: A community hospital experience. Am J Obstet Gynecol 1999;180:1364–72.
14. Jaro MA. Probabilistic linkage of large public health data files. Stat Med 1995;14:491–8.
15. Green DC, Moore JM, Adams MM, Berg CJ, Wilcox LS, McCarthy BJ. Are we understanding rates of vaginal birth after previous cesarean birth? The validity of delivery methods from birth certificates. Am J Epidemiol 1998;147:581–6.
16. Greenland S. Modeling and variable selection in epidemiologic analysis. Am J Public Health 1989;79:340–9.
17. Notzon FC. International differences in the use of obstetric interventions. JAMA 1990;263:3286–91.
18. Notzon FC, Cnattingius S, Bergsjo P, Cole S, Taffel S, Irgens L, et al. Cesarean delivery delivery in the 1980s: International comparison by indication. Am J Obstet Gynecol 1994;170:495–504.
19. Myers SA, Gleicher N. A successful program to lower cesarean delivery rates. N Engl J Med 1988;319:1511–6.
20. Sanchez-Ramos L, Kaunitz AM, Peterson HB, Martinez-Schnell B, Thompson RJ. Reducing cesarean sections at a teaching hospital. Am J Obstet Gynecol 1990;163:1081–7.
21. Socol ML, Garcia PM, Peaceman AM, Dooley SL. Reducing cesarean births at a primarily private university hospital. Am J Obstet Gynecol 1993;168:1748–54.
22. Lagrew DC, Morgan MA. Decreasing the cesarean section rate in a private hospital: Success without mandated clinical changes. Am J Obstet Gynecol 1996;174:184–91.
23. Shiono PH, McNellis D, Rhoads GG. Reasons for the rising cesarean delivery rates: 1978–1984. Obstet Gynecol 1987;69:696–700.
24. Joffe M, Chapple J, Paterson C, Beard RW. What is the optimal cesarean section rate? An outcome based study of existing variation. J Epidemiol Community Health 1994; 48:406–11.
25. Towner D, Castro MA, Eby-Wilkens E, Gilbert WM. Effect of mode of delivery in nulliparous women on neonatal intracranial injury. N Engl J Med 1999;341:1709–14.
26. Gregory KD, Korst LM, Cane P, Platt LD, Kahn K. Vaginal birth after cesarean and uterine rupture rates in California. Obstet Gynecol 1999;94:985–9.
27. Lydon-Rochelle M, Holt VL, Easterling TR, Martin DP. Risk of uterine rupture during labor among women with a prior cesarean delivery. N Engl J Med 2001;345:3–8.
28. McMahon, MJ, Luther ER, Bowes WA, Olshan AF. Comparison of a trial of labor with an elective second cesarean section. N Engl J Med 1996;335:689–95.
© 2003 The American College of Obstetricians and Gynecologists
29. Henry OA, Gregory KD, Hobel CJ, Platt LD. Using ICD-9 codes to identify indications for primary and repeat cesarean sections: Agreement with clinical records. Am J Public Health 1995;85:1143–6.