Cesarean delivery rates have increased from 20% to 33% over the past 20 years.1 These increased rates have not correlated with improved outcomes and suggest a potential overuse of this procedure. Cesarean delivery, when compared with vaginal delivery, has been associated with increased maternal morbidity, including hemorrhage, infection, and hysterectomy, and increased readmissions.2–4 Cesarean delivery rate, particularly in women at low risk, is now utilized as a reportable quality metric, and marked variation among institutions has been reported.5–7
The associations between provider volume and patient outcomes have been demonstrated across multiple medical specialties, including gynecology.8–14 Similar to surgical fields, obstetrics is procedural-based and patient outcomes are intimately linked with an obstetrician's technical skills and clinical judgment. A patient's likelihood of a cesarean delivery has been shown to be affected by individual obstetrician management decisions, such as offering induction of labor or an operative delivery.15
Few studies have examined volume relationships in obstetrics, and there are no reports of obstetrician volume on a patient's risk of cesarean delivery.16–19 Our hypothesis is that obstetricians who have higher delivery volumes may have management styles that decrease a patient's risk for cesarean delivery. In this study, we sought to answer three questions. First, do obstetricians with higher delivery volume have lower rates of cesarean delivery? Second, after adjusting for patient and obstetrician characteristics, is delivery volume an independent predictor of cesarean delivery? Third, how do the effects of delivery volume compare with the effects of an obstetrician's experience level?
MATERIALS AND METHODS
All birth records from Brigham and Women's Hospital (Boston, Massachusetts) were collected from 1 January 2000, through 31 December 2012, from the hospital's internal labor and delivery database. The database captures maternal, neonatal, labor, and delivery information for all patients admitted to the labor and delivery unit and is a compilation of information from three sources: auto-populated from hospital registration records (eg, age, race), entered by nurses at the time of admission and after delivery (eg, parity, Apgar scores, birth weight), and entered by physicians after delivery (eg, indications for induction, delivery complications, delivering obstetrician). The database fields used in this analysis were not modified or adjusted, allowing for consistent comparisons over time. All fields of the database were examined for erroneous entries by the authors (eg, gestational ages more than 44 weeks); entries were corrected, when possible, using the patient's medical record. If an entry was deemed erroneous and could not be corrected, the entry was excluded from the analysis. Furthermore, all incomplete entries were excluded from the analysis. In this 13-year period, information for 114,242 newborns delivered was in the database, of whom 112,363 were born alive and at a viable gestational age (defined as 24 weeks of gestation or more).
There are different groups of obstetricians that deliver patients in the labor and delivery unit, including an academic faculty practice and multiple private practices. In the faculty service, both specialty-trained (maternal-fetal medicine physicians) and generalist obstetricians oversee the management of patients with intimate involvement by fellow and resident trainees. Trainees do not participate in the management of private practice patients. Furthermore, there are midwifery services that practice under both the faculty and private practice groups. Because the midwives manage their patients' prenatal care and labor courses independently, we excluded all midwifery patients, whether these patients ultimately delivered by vaginal or cesarean delivery. The cesarean deliveries performed by obstetricians for midwifery patients were not included in an obstetrician's yearly delivery volume or in the calculation of a patient's risk of cesarean delivery.
The primary outcome was unscheduled cesarean delivery in labor. We considered a patient to be in labor if she either delivered vaginally or had a cesarean delivery that was coded as “unscheduled” or “emergent” for the indications of “nonreassuring fetal status” or “failure to progress,” which accounted for nearly all the cesarean deliveries performed during labor. Of the 84,122 total physician deliveries, 19,456 cesarean deliveries were excluded for not being performed in the context of labor and 6,428 deliveries were excluded for having incomplete database entries; our final sample included 58,328 patients in labor who had complete database entries.
We measured the associations between obstetrician delivery volume, experience, and unscheduled cesarean delivery using a multivariate logistic regression. We also assessed the associations by calculating adjusted cesarean delivery rates using the least squares means method. This outcome was examined in the subset of nulliparous patients with a term (37 weeks of gestation or more), singleton, vertex-presenting fetus, a standard patient group for quality reporting in obstetrics; there were 27,485 patients included in this subset.
An obstetrician's annual delivery volume was calculated by summing the number of viable, liveborn newborns delivered during a calendar year. Multiple gestation pregnancies were counted as one delivery. All types of deliveries (spontaneous vaginal, operative, and cesarean) were included to calculate an obstetrician's annual delivery volume. Obstetricians were then assigned an annual delivery volume quartile based on their number of deliveries relative to other obstetricians in the overall sample. An obstetrician's number of years of experience was calculated by subtracting the delivery year from the year when the obstetrician graduated residency. Because the experience variable was confounded by time and because cesarean delivery rates have increased over time, quartiles were also assigned based on an obstetrician's years of experience relative to other obstetricians in the overall sample.
We used a multivariate logistic regression to model the associations of annual delivery volume and obstetrician experience on a patient's risk of a cesarean delivery using all available confounders. The model was adjusted for the following variables: delivery year; patient age; patient race; parity; gestational age; induction status; induction indications (hypertension, diabetes, intrauterine growth restriction, postdates, and spontaneous rupture of membranes); use of oxytocin; birth weight; and faculty practice compared with private practice. The model was constructed with and without interaction terms between obstetrician experience and obstetrician volume based on the assumption that an obstetrician's annual volume may be affected by the number of years practiced. To test for effect modification, a likelihood ratio test was performed comparing models with and without the interaction term; P<.001, indicating the model with the interaction terms was a significantly better fit, and thus the interaction term was included in the final model. Odds ratios (ORs) and 95% confidence intervals (CIs) are reported for cesarean deliveries in all laboring patients and in the subset of nulliparous patients with term, singleton, vertex-presenting fetuses during labor.
We also calculated adjusted cesarean delivery rates for each delivery volume quartile and experience quartile. Adjusted cesarean delivery rates were derived using the least squares mean method, accounting for the same confounders as used in the regression analysis. Cesarean delivery rates are reported for all cesarean deliveries in labor and cesarean deliveries in the subset of nulliparous patients with term, singleton, vertex-presenting fetuses during labor.
Partner's Healthcare's institutional review board approved this study. Statistical analyses were performed using Stata/SE and SAS software.
Of all cesarean deliveries during this time period in the study population (n=16,194), 65.7% were performed during labor, as indicated by the indications of “nonreassuring fetal heart tracing” and “failure to progress” and were included in this analysis. Of the primary cesarean deliveries (n=10,312), these indications accounted for 78.8%.
Figure 1 compares the hospital's overall cesarean delivery rates with the cesarean delivery rates in the United States during the study time period. The hospital's cesarean delivery rate increase mirrors the national trend, increasing from 25% to 33%. Within the hospital, the cesarean delivery rate for all patients in labor increased from approximately 15% to 20%. The cesarean delivery rate in the subset of nulliparous patients with term, singleton, vertex-presenting fetuses is also shown in Figure 1, which demonstrates the increased rates of cesarean delivery in nulliparous patients in labor compared with all patients in labor, which includes multiparous patients. The cesarean delivery rate in the subset of nulliparous patients with term, singleton, vertex-presenting fetuses fluctuated between 20% and 30% over this time period.
The patient and obstetrician characteristics for each delivery volume quartile are shown in Table 1. Because the patients were assigned to groups based on their obstetrician's delivery volume quartile, the quartiles contain an increasing number of patients. The quartile groups were compared for each characteristic, and all were found to be statistically different (P<.001). These characteristics were included in the analyses.
Table 2 displays the ranges for the quartiles of delivery volume (in number of deliveries per year) and experience (number of years between completing residency and the delivery year). The median number of deliveries per year per obstetrician was 60. The lowest-volume obstetricians, quartile 1, performed fewer than 33 deliveries per year. The highest-volume obstetricians, quartile 4, performed at least 130 deliveries per year. The median number of years of experience was 11. The least experienced obstetricians, quartile 1, had been practicing for up to 4 years. The most experienced, quartile 4, had been in practice for at least 19 years. The data for volume and experience were both right-skewed, with the fourth quartile having the largest intraquartile range of deliveries (130–412) and years of experience (19–40). Of the 169 obstetricians, 47 (28%) did not change volume quartile over their years in practice, 60 (36%) changed by one quartile, 42 (25%) changed by two quartiles, and 20 (12%) changed by three quartiles. The most common scenario for the obstetricians who changed two or more quartiles was decreasing volume over time.
The associations between obstetrician delivery volume and experience as derived from the multivariable logistic regression are shown in Table 3. The ORs and 95% CIs are shown for all cesarean deliveries during labor and for the subset of nulliparous patients with term, singleton, vertex-presenting fetuses. The results from the regression were notable for an OR more than 2 for cesarean delivery among patients delivered by obstetricians who performed fewer than the median deliveries per year: OR was 2.00 for quartile 1 (95% CI 1.68–2.38) and OR was 2.73 for quartile 2 (95% CI 2.40–3.11) as compared with quartile 4 as a reference. In contrast, obstetrician experience had a smaller effect on a patient's odds of cesarean delivery and was only significant for quartile 1. The associations with volume were similar for the subset of nulliparous patients with term, singleton, vertex-presenting fetuses.
The ORs and CIs for the other terms in the model are included in Table 3. Delivery year was included as a model term because the cesarean delivery rate increased over time, consistent with national trends. Increasing gestational age was associated with a decreased OR in the overall model (OR 0.92); however, in term patients, increasing gestational age was associated with an increased OR in the subset of nulliparous patients with term, singleton, vertex-presenting fetuses (OR 1.21), likely because of the effects of postterm patients. In addition, those factors with notable increased odds of cesarean delivery in the overall model included the following: parity (OR 6.15 for nulliparity compared with parity 2 or more); self-identified race of “black” and “Hispanic” (OR 2.43 and 2.02, respectively); oxytocin use (OR 1.18); and being a patient affiliated with the faculty practice (OR 1.28). Undergoing an induction of labor was not associated with an increased odds of cesarean delivery in the overall model (OR 0.97, 95% CI 0.89–1.05); however, it was associated with increased odds in the subset of nulliparous patients with term, singleton, vertex-presenting fetuses (OR 1.37). The models included an interaction term between obstetrician volume and experience to account for any effects of experience on volume; the effects of these interactions are not listed in Table 3. The interaction term and all the variables shown in Table 3 were included in the final model; thus, the effects of obstetrician volume are independent of all these variables.
The adjusted cesarean delivery rates based on obstetrician volume and experience are shown in Figure 2. Rates are shown for patients during labor and in the subset of nulliparous patients with term, singleton, vertex-presenting fetuses during labor. After adjustment, the cesarean delivery rate for all laboring patients was markedly increased for obstetricians who performed fewer than the median number of deliveries: 18.2% and 20.0% for quartiles 1 and 2 compared with 12.4% and 9.2% for quartiles 3 and 4 (P<.001). In comparison, the cesarean delivery rates were clinically similar, although statistically different, across all experience quartiles: 15.9%, 15.7%, 14.0%, and 14.2% (P<.001). Similar trends were seen in the subset of nulliparous patients with term, singleton, vertex-presenting fetuses. The adjusted rates in the subset of nulliparous patients with term, singleton, vertex-presenting fetuses are higher given the associated risk for cesarean delivery because of the effects of nulliparity.
Two additional secondary analyses were performed to further understand specific obstetrician characteristics that may be affecting the model. The first subset analysis focused on physicians who changed from a high-volume (quartile 3 or 4) to low-volume provider (quartile 1 or 2) during the study period. Patients delivered by these obstetricians during their low-volume years had a higher risk of cesarean delivery compared with those delivered by high-volume obstetricians: OR was 1.66 (95% CI 1.28–2.15) for quartile 1, OR was 2.10 (95% CI 1.73–2.56) for quartile 2, and OR was 1.16 (95% CI 0.95–1.41) for quartile 3 when compared with quartile 4. We also examined the private practice groups only. Patients delivered by low-volume private practice obstetricians had an increased risk of cesarean delivery: OR was 2.7 (95% CI 2.14–3.28) for quartile 1, OR was 3.4 (95% CI 2.97–4.00) for quartile 2, and OR was 1.13 (95% CI 0.97–1.31) as compared with quartile 4.
Obstetrician delivery volume is a potentially modifiable risk factor for cesarean delivery. Patients delivered by obstetricians with the lowest delivery volumes (fewer than the median of 60 deliveries per year) had a twofold increased rate of cesarean delivery as compared with patients delivered by higher-volume obstetricians, independent of other patient and obstetrician characteristics. In contrast, obstetrician experience had a smaller effect. These findings support our hypothesis that managing labor frequently may affect an obstetrician's likelihood to perform a cesarean delivery. The highest-volume obstetricians, not the most experienced, have cesarean delivery rates below those of their peers and appear to have different practice patterns than low-volume obstetricians (eg, higher rates of induction, higher use of oxytocin, and higher operative vaginal delivery rates).
Our results revealed that volume quartile 2 had an increased OR for cesarean delivery as compared with quartile 1, although the ORs for both quartile 1 and quartile 2 were significantly higher than those of the other quartiles. It is possible that the obstetricians in the lowest quartile practice too infrequently to appreciate the true associations of their practice patterns on labor outcomes. Also, the sample size of quartile 1 (n=2,949) is significantly smaller than those of quartiles 2 through 4 (n=7,576, n=15,307, and n=32,496, respectively), which may affect the magnitude of OR for quartile 1. The discrepancy in quartile sizes is attributed to the patient assignment based on obstetrician volume.
Some obstetricians did change volume quartiles over time, primarily from high to low volume over the course of their career. However, in the secondary analysis of low-volume obstetricians who had been considered high-volume (quartile 3 or 4) during another year, the patients of those obstetricians were at similarly increased risk for cesarean delivery. Thus, it appears that the increased risk of cesarean delivery is related to an obstetrician's current volume and is not significantly modified by the volumes of past years.
This study is not without limitations. The practice patterns at this institution may not be reflective of other hospitals.20 In examining the sample patient characteristics, the maternal age is older and the preterm birth rate is higher compared with national averages.21 We also performed a secondary analysis looking at only private practitioners and excluding the teaching service, because this may be more applicable to the majority of deliveries performed in the United States; the ORs were even higher in this subgroup. Second, regarding residual confounding, we accounted for all known differences in patient case mix in the multivariate analysis. However, proxies for comorbidities were coded as induction indications and are likely underreported in the sample. There may also be some residual differences in patient mix that we were unable to account for; however, when comparing a similar patient population across quartiles (spontaneous, term, singleton deliveries), the relationships with volume and experience remain significant. Finally, the model treats all practice groups equally; however, their group size, call schedules, and reimbursement structures vary. This study is unable to assess the effects of practice-specific characteristics.
Obtaining patient-level and physician-level data from a single institution allows for controlling for both patient and physician characteristics, which is a limitation of population-based, de-identified administrative databases and a strength of this analysis. Although the design of this study did not allow for the calculation of specific volume thresholds, ORs and cesarean delivery rates were dramatically higher for obstetricians with fewer than the median number of deliveries per year at our institution (60 deliveries per year). The findings of this study may prompt further discussion regarding the role for procedural volume as a component of credentialing at institutions, for state licensing, and for board certification. Furthermore, these findings suggest that practice models that spread delivery volume over a large number of obstetricians may result in increased cesarean delivery rates compared with practice models that direct patients toward high-volume obstetricians, such as a laborist model. Given the limitations of studying a single institution in an obstetrician-dense area, additional research is needed to further explore the effects of volume on patient outcomes.
2. Ye J, Betrán AP, Vela MG, Souza JP, Zhang J. Searching for the optimal rate of medically necessary cesarean delivery. Birth 2014. DOI: 10.1111/birt.12104 [Epub ahead of print].
3. Liu S, Heaman M, Joseph KS, Liston RM, Huang L, Sauve R, et al.. Risk of maternal postpartum readmission associated with mode of delivery. Obstet Gynecol 2005;105:836–42.
4. Liu S, Liston RM, Joseph KS, Heaman M, Sauve R, Kramer MS. Maternal mortality and severe morbidity associated with low-risk planned cesarean delivery versus planned vaginal delivery at term. Can Med Assoc J 2007;176:455–60.
5. Janakiraman V, Ecker J. Quality in obstetric care: measuring what matters. Obstet Gynecol 2010;116:728–32.
6. Main EK, Moore D, Farrell B, Schimmel LD, Altman RJ, Abrahams C, et al.. Is there a useful cesarean birth measure? Assessment of the nulliparous term singleton vertex cesarean birth rate as a tool for obstetric quality improvement. Am J Obstet Gynecol 2006;194:1644–51.
7. Srinivas SK, Fager C, Lorch SA. Evaluating risk-adjusted cesarean delivery rate as a measure of obstetric quality. Obstet Gynecol 2010;115:1007–13.
8. Birkmeyer JD, Stukel TA, Siewers AE, Goodney PP, Wennberg DE, Lucas FL. Surgeon volume and operative mortality in the United States. N Engl J Med 2003;349:2117–27.
9. Ho V, Heslin MJ, Yun H, Howard L. Trends in hospital and surgeon volume and operative mortality for cancer surgery. Ann Surg Oncol 2006;13:851–8.
10. Cowan JA Jr, Dimick JB, Thompson BG, Stanley JC, Upchurch GR Jr. Surgeon volume as an indicator of outcomes after carotid endarterectomy: an effect independent of specialty practice and hospital volume. J Am Coll Surg 2002;195:814–21.
11. Halm EA, Lee C, Chassin MR. Is volume related to outcome in health care? A systematic review and methodologic critique of the literature. Ann Intern Med 2002;137:511–20.
12. Chowdhury MM, Dagash H, Pierro A. A systematic review of the impact of volume of surgery and specialization on patient outcome. Br J Surg 2007;94:145–61.
13. Boyd LR, Novetsky AP, Curtin JP. Effect of surgical volume on route of hysterectomy and short-term morbidity. Obstet Gynecol 2010;116:909–15.
14. Wallenstein MR, Ananth CV, Kim JH, Burke WM, Hershman DL, Lewin SN, et al.. Effect of surgical volume on outcomes for laparoscopic hysterectomy for benign indications. Obstet Gynecol 2012;119:709–16.
15. Spong CY, Berghella V, Wenstrom KD, Mercer BM, Saade GR. Preventing the first cesarean delivery: Summary of a Joint Eunice Kennedy Shriver
National Institute of Child Health and Human Development, Society for Maternal-Fetal Medicine, and American College of Obstetricians and Gynecologists Workshop. Obstet Gynecol 2012;120:1181–93.
16. Janakiraman V, Lazar J, Joynt KE, Jha AK. Hospital volume, provider volume, and complications after childbirth in U.S. hospitals. Obstet Gynecol 2011;118:521–7.
17. Snowden JM, Darney BG, Cheng YW, McConnell KJ, Caughey AB. Systems factors in obstetric care: the role of daily obstetric volume. Obstet Gynecol 2013;122:851–7.
18. Snowden JM, Cheng YW, Kontgis CP, Caughey AB. The association between hospital obstetric volume and perinatal outcomes in California. Am J Obstet Gynecol 2012;207:478.e1–7.
19. Klein MC, Spence A, Kaczorowski J, Kelly A, Grzybowski S. Does delivery volume of family physicians predict maternal and newborn outcome? CMAJ 2002;166:1257–63.
20. Baicker K, Buckles KS, Chandra A. Geographic variation in the appropriate use of cesarean delivery. Health Aff (Millwood) 2006;25:w355–67.