Kahn, Emily B. PhD, MPH; Berg, Cynthia J. MD, MPH; Callaghan, William M. MD, MPH
In 2005, the United States reported its highest-ever rate of cesarean delivery (30.3% of all U.S. births).1 This high rate has been attributed to both an increase in the rate of primary cesarean deliveries from 1996 to 20042 (14.6% to 20.6%) and a steep decline in the rate of vaginal birth after cesarean delivery during the same time period (28.3% to 9.2% of women delivering after a previous cesarean delivery).3 Given the decline in vaginal birth after cesarean, current efforts to reduce the cesarean rate have focused on reducing the occurrence of primary cesarean delivery among women with low-risk pregnancies.
During the past 10 years, a number of studies have used birth certificate data to provide population-based estimates of the rate of “unnecessary” cesarean delivery.4–7 More recently, birth certificate data have been used to examine the primary cesarean delivery rate among women with “no indicated risk.”8 The construct of no-indicated-risk cesarean births in this context is defined as those for whom the pregnancy is low risk, ie, the fetus is full-term, singleton, and in the vertex position, and the mother has no reported medical risk factors or complications of labor and/or delivery identified on the birth certificate.
Although using birth certificate data to estimate the frequency of no-indicated-risk deliveries has the advantage that the data are readily available on a national basis, previous studies9–13 have shown consistent underreporting of many risk factors and complications on the birth certificate. Moreover, administrative hospital discharge data have been shown to be more sensitive than birth certificates for identifying pregnancy risk factors and labor complications.11 Using linked birth certificate–hospital discharge data from Georgia, we explored the effect of using birth certificate data alone to estimate the proportion of cesarean deliveries that were no indicated risk and examined the agreement between the two data sources on the presence of risk factors. We hypothesized that hospital discharge data would identify indicators of maternal risk when the matched birth certificate did not indicate such risk.
MATERIALS AND METHODS
The Georgia Division of Public Health’s Office of Health Information and Policy provided an electronic database that linked birth certificates with discharges for delivery hospitalizations from the statewide hospital discharge database from 1999 to 2004, with a common anonymous identifier generated by the Office of Health Information and Policy from a combination of characters from the woman’s first and last names, date of birth, and sex.
The discharge database is an administrative billing data set that includes information on discharges from all 89 civilian hospitals in Georgia, with up to 10 International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes for each hospitalization. The hospital discharge database was used to determine the mother’s hospital admission and discharge dates and to identify ICD-9-CM codes for risk factors and complications. The birth certificate was used to ascertain race, ethnicity, and age of mother, plurality and gestational age of the newborn, method of delivery, and medical risk factors during pregnancy and complications of labor and delivery (hereafter referred to as risk factors and complications).
Between January 1, 1999, and December 31, 2004, 804,601 women delivered liveborn neonates (Fig. 1). Of these deliveries, 82% (662,896) were defined from information on birth certificates as meeting the Healthy People 2010 definition of low risk—delivery of a singleton neonate in the vertex position with gestational age between 37 and 41 weeks—and were retained for further analyses. Using the Office of Health Information and Policy’s linking variable, 85.3% (565,767) of birth certificates for low-risk pregnancies were linked to their hospital discharge records. Once all birth certificate and hospital discharge information for a woman was linked, the delivery hospitalization was selected as the one in which the newborn’s birth date was between the mother’s admission date and discharge date. Of 565,767 women with low-risk pregnancies whose data were successfully linked, 12.4% (70,255) had primary cesarean deliveries according to the birth certificate. The authors received a de-identified data file from the state of Georgia; hence, the study was exempt from review by the Centers for Disease Control and Prevention’s institutional review board.
The construct of “no indicated risk” in the literature uses the 31 risks and complications listed on the birth certificate as indicators of potential risk for the woman. Although such risk factors might make cesarean delivery more likely, they are not necessarily indications for the procedure. We were able to match 26 of the birth certificate indicators to ICD-9-CM codes in the hospital discharge data set; this group of ICD-9-CM codes made up Risk Definition 1 (Table 1) and mirrors the definition used in previous literature. Maternal seizures during labor were presumed to have been captured by the ICD-9-CM codes for eclampsia and were not mapped to additional codes because there are no specific codes for seizures during labor. Codes for the remaining four risk factors (previous neonate 4,000 grams or larger, previous preterm/small for gestational age neonate, and rubella or syphilis complicating the current pregnancy) do not exist within ICD-9-CM and thus were not included in Risk Definition 1. To be consistent with earlier studies, we used Risk Definition 1 and defined a no-indicated-risk cesarean delivery using only birth certificate data to identify primary cesarean deliveries to women with low-risk pregnancies for which the birth certificate indicated no risk factors or complications.
We also created a second risk definition in an attempt to establish a set of risk factors that were more likely to be directly related to the indication for cesarean delivery (Risk Definition 2) (Table 1). This definition included all risk factors and complications from the birth certificate except anemia and precipitous labor, which were removed because they are unlikely to necessitate cesarean delivery. Based on clinical understanding of the reasons for cesarean delivery, the following selected conditions, available in the hospital discharge data, were added: repeat cesarean delivery, other uterine scar, macrosomia affecting maternal management, prolonged rupture of membranes, abnormalities of organs and soft tissues of pelvis, infection of the amniotic cavity, intrauterine growth restriction affecting maternal management, and known or suspected fetal abnormality affecting management of the mother.
Among women who underwent no-indicated-risk cesarean deliveries based on birth certificate data only, we created a count variable to indicate the number of Risk Definition 1 conditions whose ICD-9-CM codes were recorded on that hospital discharge record. This variable was used to calculate the percentage of no-indicated-risk cesarean deliveries based on birth certificate data only that had one or more of the risk factors or complications reported in the hospital discharge data.
To examine the agreement between the birth certificate and the hospital discharge data, we compared risk factors and complications reported in the two data sources for all 565,757 low-risk deliveries with linked data. We calculated overall agreement and agreement for each risk factor or complication between the two data sources. To assess the statistical significance, we used the kappa statistic, which measures the extent to which the agreement between two or more data sources exceeds the level that would be expected by chance alone.
Finally, we used both Risk Definitions 1 and 2 and combinations of birth certificate and hospital discharge data to calculate the percentages of low-risk linked primary cesarean deliveries with no indicated risk. To compare the rates of no-indicated-risk cesarean deliveries in our data with previously published rates, we also calculated cesarean delivery rates among low-risk pregnancies in which there was no indicated risk using both risk definitions and data sources. All analyses were conducted with SAS 9.1 (SAS Institute, Cary, NC).
The age distribution for women with linked data was similar to that for women with unlinked data, and similar proportions were primiparous (12.4% and 10.9%, respectively) and had primary cesarean deliveries (40.6% compared with 41.6%). However, women whose deliveries did not link to hospital discharge records were more likely to be of other race (5.2% compared with 3.1%) or Hispanic ethnicity (27.0% compared with 9.7%). Women with low-risk pregnancies whose deliveries linked were also more likely to have risk factors on the birth certificate (20.9%) than were women whose deliveries did not link (18.7%). Of the 565,767 women with low-risk pregnancies with linked data, 70,255 (12.4%) had primary cesarean deliveries. Of these, 40,932 (58.3%) had no risk factors or complications reported on the birth certificate and thus comprised the no-indicated-risk cesarean delivery based on birth certificate data only (NIR-BC) group.
Overall, 87.4% of the women in the NIR-BC group had at least one Risk Definition 1 condition code in the hospital discharge data (Table 2). The most common risk among no-indicated-risk primary cesarean deliveries based on birth certificate data only was dysfunctional labor (36.8%), followed by fetal distress or abnormal fetal heart rhythm (30.8%), cephalopelvic disproportion (18.3%), anemia (12.9%), and hypertension (11.6%). Malpresentation was indicated in 10.4% of the hospital discharge records, even though breech presentation was an exclusion criterion for the low-risk group. Similarly, despite being reported as primary cesarean deliveries on the birth certificate, 6.4% of the no-indicated-risk cesarean deliveries based on birth certificate data only had an ICD-9-CM code in the hospital discharge record indicating a previous cesarean delivery.
The overall agreement between data sources for the presence of any risk factor or complication was low (κ=0.18). Among individual risk factors or complications, kappa statistics ranged from moderate (κ=0.44) for diabetes to very low (κ=0.01) for hemoglobinopathies; only 8 of 25 risk factors or complications had a kappa statistic above 0.2. In addition to diabetes, placenta previa, hypertension, placental abruption, cord prolapse, and cephalopelvic disproportion had the highest kappa statistics (ranging from 0.39 to 0.36), but fewer than one third of the women with these risks indicated in the hospital discharge data had them listed on the birth certificate (data not shown).
Dysfunctional labor, which according to the hospital discharge data was the most prevalent risk factor among women with no-indicated-risk cesarean deliveries based on birth certificate data only, had low agreement between the birth certificate and hospital discharge record (κ=0.08). Only 5.2% of women with ICD-9-CM codes indicating dysfunctional labor had this problem indicated on the birth certificate (data not shown).
Using birth certificate data alone, 58.3% of primary cesarean deliveries among linked low-risk pregnancies had no indicated risk under Risk Definition 1, which includes all birth certificate risk factors and conditions; 59.2% had no indicated risk using Risk Definition 2, which omits precipitous labor and anemia as risks for cesarean delivery (Table 3). However, only 7.4% and 3.9% of primary cesarean deliveries among linked low-risk pregnancies had no indicated risk under Risk Definitions 1 and 2, respectively, when information was used from both birth certificates and hospital discharge data. The percentage of cesarean deliveries among low-risk pregnancies that had no indicated risk was less than 10% for all combinations of risk definitions and data sources when hospital discharge data were included as a data source.
Finally, we calculated the primary cesarean delivery rates among linked low-risk pregnancies with no risk factors or complications by risk definition and data source (Table 4). When only birth certificate data were used to identify risk factors, the primary cesarean delivery rate was 9.0–9.1% among these women. When hospital discharge data were also used to identify risk factors and complications, the primary cesarean delivery rate among women with low-risk pregnancies was 1.2–2.0%.
Judging from our analysis of live-birth deliveries in Georgia, birth certificate and hospital discharge data are likely to provide very different estimates of the percentage of women with low-risk pregnancies with a no-indicated-risk cesarean delivery and of the cesarean delivery rate among women with low-risk pregnancies with no indicated risk. Using population-based, linked birth certificate–hospital discharge data for low-risk pregnancies, we found that 87% of cesarean deliveries with no indicated risk for cesarean on the birth certificate did, in fact, have a risk factor or complication in the linked hospital discharge data. All risk factors and complications, including those with high prevalence, were underreported on the birth certificate as compared with the hospital discharge data, and there was little agreement between the two sources. Using birth certificate data alone, we would have estimated that 9% of women with low-risk pregnancies and no indicated risk had cesarean deliveries. By adding hospital discharge data, the cesarean delivery rate among women with low-risk pregnancies with no indicated risk was reduced fourfold to sevenfold to 1–2%.
Although birth certificates are widely available and contain information on medical risk factors and complications of labor and delivery, our study indicates that sole reliance on birth certificate reporting led to overestimation of the proportion of women with low-risk pregnancies who had primary cesarean deliveries with no indicated risk. Although specificity for risk factors and conditions on birth certificates is generally fair to good, sensitivity is generally poor.10 Because some complications that occur during labor and delivery, such as dysfunctional labor, are sometimes indicated on the birth certificate and not in hospital discharge data (data not shown), the most comprehensive method to identify women who have no indicated risk is to look for risk factors and complications in both hospital discharge and birth certificate data.
The increasing rate of no-indicated-risk cesarean delivery over time implies that more such cesarean deliveries are occurring, that reporting of risk factors or complications on birth certificates is poorer, or that risk factors have emerged that are not included on birth certificates. Still, the no-indicated-risk cesarean delivery rate among women with low-risk pregnancies that we found when using only birth certificates to identify risk was six times that found when hospital discharge data were used to identify risk (9.1% compared with 1.4%).
The birth certificate is designed to collect information on a wide variety of risk factors and conditions that may result in a poor birth outcome. It was not designed to collect information on specific risks or indications for cesarean delivery. The birth certificate’s poor predictive value in identifying risk factors and complications has been well documented,9,11,13–18 and hospital discharge data have been demonstrated to reflect maternal medical conditions with greater validity.11
This study is subject to some limitations. Overall, 15% of births in our data set did not link to a hospital discharge record, with Hispanic women and women of “other” race being more likely to be in the unlinked group. However, this would introduce bias only if the reporting of risk on birth certificates and in hospital discharge databases differed by race and ethnicity. We compared reporting of risk factors on birth certificates that linked to hospital discharge records and on those that did not link. There are no systematic differences; some risk factors were reported more frequently in the linked group and some in the unlinked group. For all risk factors, differences were in the tenths of a percent range. Also, neither data source provides information on other key risk factors for cesarean delivery, such as maternal obesity. In addition, some of the risk factors and complications on the birth certificate do not have direct correlates in the hospital discharge data. For example, “other excessive bleeding” is listed as a complication of labor and delivery on the birth certificate, but no information is provided on the birth certificate about whether the bleeding occurred intrapartum or postpartum. Because postpartum hemorrhage would not be a reason to perform a cesarean delivery, we used ICD-9-CM codes for antepartum hemorrhage not due to placental problems to indicate excessive bleeding. Finally, our data came from a single state rather than the entire country. Although studies in other states have shown a similar lack of concordance between birth certificate and hospital discharge data with regard to risk factors and complications,11 there is substantial variability among states in the completeness of reporting of birth certificate items, and results from this state could differ from results that used data from a different state.
The debate over the appropriate use of cesarean delivery has been ongoing for more than 100 years, and concerns about rising rates of cesarean delivery have been voiced since the 1980s.19 Many factors other than medical necessity have been cited as playing a role in the increase, including patient request, reduced exposure during physician training to operative vaginal delivery, and increased concerns over liability and malpractice.19–21 Advances in surgical and anesthetic techniques have contributed to dramatic reductions in both maternal and infant mortality related to cesarean delivery. At the same time, it is undeniable that cesarean deliveries have costs. Data regarding the incremental benefits of cesarean delivery to maternal and neonatal health are inconclusive; the risks attendant to any major surgical procedure, the increased risk for women who have pregnancies after cesarean delivery, and the poor predictive value of electronic fetal heart rate monitoring resulting in intervention with little benefit all have been noted.2 Moreover, a recent conservative estimate places the financial burden of cesarean delivery in the range of $15 billion a year.
There is a lack of clarity in the clinical and research communities with attempts to explain the rise in primary cesarean delivery, particularly those that appear to occur without indication of medical risk. Even within geographic regions, the variation in cesarean delivery rates among institutions suggests a pattern of random decision making.23 Linked data, such as the birth certificate–hospital discharge data sets, can provide more information than either source alone, and they may be used to examine risk for other poor birth outcomes in addition to reasons for cesarean delivery. However, in exploring reasons behind the increase in cesarean delivery without medical risk, it is important to keep in mind that indicators of risk tell us little about the decision-making process involved in performing a cesarean delivery. To understand the dynamics of the clinical decision making and the interactions between women and their physicians, it will be important to move beyond information found in vital records and administrative data sets. In the absence of more nuanced information, the construct of no indicated risk as determined from birth certificates should be interpreted with caution, and the use of linked data should be considered whenever possible.
1. Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Menacker F, Kirmeyer S, et al. Births: final data for 2005. Natl Vital Stat Rep 2007;56:1–103.
2. Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Menacker F, Kirmeyer S. Births: final data for 2004. Natl Vital Stat Rep 2006;55:1–101.
3. Menacker F, Declercq E, Macdorman MF. Cesarean delivery: background, trends, and epidemiology. Semin Perinatol 2006;30:235–41.
4. Koroukian SM, Trisel B, Rimm AA. Estimating the proportion of unnecessary Cesarean sections in Ohio using birth certificate data [published erratum appears in J Clin Epidemiol 1999;52:379]. J Clin Epidemiol 1998;51:1327–34.
5. Kabir AA, Steinmann WC, Myers L, Khan MM, Herrera EA, Yu S, et al. Unnecessary cesarean delivery in Louisiana: an analysis of birth certificate data. Am J Obstet Gynecol 2004;190:10–9.
6. Koroukian SM, Rimm AA. Declining trends in cesarean deliveries, Ohio 1989-1996: an analysis by indications. Birth 2000;27:12–8.
7. Denk CE, Kruse LK, Jain NJ. Surveillance of cesarean section deliveries, New Jersey, 1999-2004. Birth 2006;33:203–9.
8. Declercq E, Menacker F, Macdorman M. Rise in “no indicated risk” primary caesareans in the United States, 1991-2001: cross sectional analysis. BMJ 2005;330:71–2.
9. DiGiuseppe DL, Aron DC, Ranbom L, Harper DL, Rosenthal GE. Reliability of birth certificate data: a multi-hospital comparison to medical records information. Matern Child Health J 2002;6:169–79.
10. Lydon-Rochelle MT, Holt VL, Cardenas V, Nelson JC, Easterling TR, Gardella C, et al. The reporting of pre-existing maternal medical conditions and complications of pregnancy on birth certificates and in hospital discharge data. Am J Obstet Gynecol 2005;193:125–34.
11. Lydon-Rochelle MT, Holt VL, Nelson JC, Cardenas V, Gardella C, Easterling TR, et al. Accuracy of reporting maternal in-hospital diagnoses and intrapartum procedures in Washington State linked birth records. Paediatr Perinat Epidemiol 2005;19:460–71.
12. Reichman NE, Hade EM. Validation of birth certificate data. A study of women in New Jersey’s HealthStart program. Ann Epidemiol 2001;11:186–93.
13. Roohan PJ, Josberger RE, Acar J, Dabir P, Feder HM, Gagliano PJ. Validation of birth certificate data in New York State. J Community Health 2003;28:335–46.
14. Northam S, Knapp TR. The reliability and validity of birth certificates. J Obstet Gynecol Neonatal Nurs 2006;35:3–12.
15. Gore DC, Chez RA, Remmel RJ, Harahan M, Mock M, Yelverton R. Unreliable medical information on birth certificates. J Reprod Med 2002;47:297–302.
16. Dobie SA, Baldwin LM, Rosenblatt RA, Fordyce MA, Andrilla CH, Hart LG. How well do birth certificates describe the pregnancies they report? The Washington State experience with low-risk pregnancies. Matern Child Health J 1998;2:145–54.
17. Piper JM, Mitchel EF Jr, Snowden M, Hall C, Adams M, Taylor P. Validation of 1989 Tennessee birth certificates using maternal and newborn hospital records. Am J Epidemiol 1993;137:758–68.
18. Zollinger TW, Przybylski MJ, Gamache RE. Reliability of Indiana birth certificate data compared to medical records. Ann Epidemiol 2006;16:1–10.
19. Cyr RM. Myth of the ideal cesarean section rate: commentary and historic perspective. Am J Obstet Gynecol 2006;194:932–6.
20. National Institutes of Health State-of-the-science conference statement: Cesarean delivery on maternal request March 27-29, 2006. Obstet Gynecol 2006;107:1386–97.
21. Kalish RB, McCullough LB, Chervenak FA. Decision-making about caesarean delivery. Lancet 2006;367:883–5.
22. Resnik R. Can a 29% cesarean delivery rate possibly be justified? Obstet Gynecol 2006;107:752–4.
23. Clark SL, Belfort MA, Hankins GD, Meyers JA, HouserFM. Variation in the rates of operative delivery in the United States. Am J Obstet Gynecol 2007;196:526.e1–5.
Figure. No caption available.
© 2009 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.