Hanley, Gillian E. MA; Janssen, Patricia A. PhD; Greyson, Devon MLIS
OBJECTIVE: To examine regional variations in rates of primary cesarean delivery and assisted vaginal delivery in the population of British Columbia, while adjusting for the maternal characteristics and conditions that increase the likelihood of operative delivery.
METHODS: Using data from the British Columbia Perinatal Database Registry, we studied all deliveries in British Columbia between 2004 and 2007, excluding women who had a previous cesarean delivery (n=116,839). Our primary outcome of interest was mode of delivery, further defined as delivery by cesarean or assisted vaginal delivery. We calculated crude and risk-adjusted rates of primary cesarean delivery and assisted vaginal delivery across British Columbia's 16 Health Service Delivery Areas and examined cesarean delivery rates by indication for the procedure.
RESULTS: Crude primary cesarean delivery and assisted vaginal delivery rates varied markedly across the Health Service Delivery Areas ranging from 16.1 to 27.5 per 100 deliveries, and from 8.6 to 18.6 per 100 deliveries, respectively. The most common indication for cesarean delivery was dystocia, which accounted for 30.0% of all cesarean deliveries and varied more than fivefold across regions. After controlling for maternal characteristics and conditions known to increase the likelihood of cesarean delivery and assisted vaginal delivery, adjusted cesarean delivery rates varied twofold, ranging from 14.7 to 27.6 per 100 deliveries, while adjusted assisted vaginal delivery rates varied by more than twofold, ranging from 6.5 to 15.3 per 100 deliveries.
CONCLUSION: Our results illustrate substantial regional variation in the use of cesarean delivery that cannot be explained by patient illness or preferences. This variation likely reflects differences in practitioners' approaches to medical decision-making.
LEVEL OF EVIDENCE: II
Canada's cesarean delivery rate has increased dramatically over the last two decades, reaching an all time high of 26.3% of in-hospital deliveries in 2005–20061; however, the recent increase in the cesarean delivery rate has not been accompanied by a corresponding positive impact on maternal or perinatal outcomes.2,3 The significantly lower rates of cesarean delivery in countries with comparable maternal and perinatal outcomes, for example Norway at 16.4% in 2006, suggests that Canada's higher cesarean delivery rate may not be necessary to achieve such outcomes.4,5 Rather, researchers have shown that, compared with those delivering vaginally, women undergoing low-risk, planned cesarean deliveries experience increased risks of cardiac arrest, wound hematoma, hysterectomy, major puerperal infection, anesthetic complication, venous thromboembolism, hemorrhage requiring hysterectomy, and a significantly increased rate of postpartum readmission to hospital.6,7 Rising cesarean delivery rates also have important economic implications. A comparison of costs per mother/neonate pair in Nova Scotia demonstrated that cesarean deliveries in labor were 60% more expensive than spontaneous vaginal delivery and 35% more expensive than assisted vaginal delivery.8 These costs often extend to subsequent deliveries, as research has shown that almost three quarters of women who had an initial cesarean delivery in labor and who went on to have more children underwent at least one repeat cesarean delivery.9
The Society of Obstetricians and Gynaecologists of Canada continues to raise concerns that the rising cesarean delivery rate reflects the use of the procedure in cases in which it is not medically indicated.10 A fundamental principle in quality assessment of health care is that regional variation diminishes as quality improves—unwarranted variation in a health care procedure is often used as an indicator of poor quality of care.11–13 Examining regional variations in cesarean delivery rates across populations is a means of assessing whether the surgical procedure is being used appropriately. This article is the first to examine small area regional variations, at the level of the Health Services Delivery Area, in rates of cesarean delivery and assisted vaginal delivery in the population of British Columbia. The objective of our study is to estimate whether adjusting for maternal characteristics and conditions that are known risk factors for cesarean delivery eliminates regional variation. Residual regional variations would indicate that nonmedical factors, such as differences in socioeconomic status, differences in provider mix in a region (eg, number of midwives, obstetricians, and family physicians) and differences in the capacity of different facilities to respond similarly to a medical emergency, were influencing the cesarean delivery rate.
MATERIALS AND METHODS
There are approximately 40,000 live births annually in British Columbia, varying across the Health Services Delivery Areas (due to differences in population size) from a low of approximately 400 to a high of approximately 5,500 annual births. This population-based cross-sectional study used anonymized data for the calendar years 2004 to 2007 from the British Columbia Perinatal Database Registry, which includes information on approximately 99% of births in the province. Validation studies revealed that the British Columbia Perinatal Database Registry is an accurate and comprehensive source of perinatal information for British Columbia with accuracy rates of 97% over all data fields and rates of missing data less than 0.01%.14 A certificate of ethics approval for this study was obtained from the Clinical Research Ethics Board at the University of British Columbia.
The British Columbia Perinatal Database Registry data indicates there were 153,750 deliveries in British Columbia during our study period. Recent obstetric practice in North America has been characterized by a shift away from planning vaginal birth after cesarean (VBAC).15 To eliminate previous cesarean births as a potential cause of regional variation, we excluded mothers with a previous cesarean delivery. All other mothers who delivered in British Columbia between 2004 and 2007 were included in our study, resulting in a study sample of 116,839. We created 16 dichotomous variables representing a mother's Health Services Delivery Area of residence—our primary exposure of interest. Our primary outcome of interest was mode of delivery, further defined as delivery by cesarean or assisted vaginal delivery. We coded all deliveries using forceps alone, forceps and vacuum, or vacuum alone as assisted vaginal delivery. To eliminate bias caused by different rates of maternal characteristics and conditions known to increase the risk of cesarean and assisted vaginal delivery, we controlled for these individual determinants. These potential determinants of primary cesarean delivery and assisted vaginal delivery were divided into two groups: maternal characteristics and maternal conditions.16 The maternal characteristics examined included maternal age, prepregnancy body mass index (BMI), weight gain during pregnancy, smoking during pregnancy, and parity. The maternal conditions we controlled for were hypertensive disorders in pregnancy, preexisting or gestational diabetes, preterm delivery (less than 37 weeks gestational age), prior fetal death, prior neonatal death, breech or transverse presentation in labor and delivery, and gestation with multiples.
We calculated overall crude rates of delivery by cesarean delivery, assisted vaginal delivery, and spontaneous vaginal delivery for each Health Services Delivery Area. To be consistent with the Canadian Perinatal Surveillance System, we calculated rates of assisted vaginal delivery after excluding women who delivered by cesarean. Because it has been suggested that induction of labor may influence the rates of operative delivery, we also presented crude rates of labor induction across the Health Services Delivery Areas. We then examined crude cesarean delivery rates by indication for the procedure across the Health Services Delivery Areas to examine whether the level of regional variation in cesarean delivery rates differed according to the indication for the procedure. We calculated regional risk-adjusted rates for both cesarean and assisted vaginal delivery using multivariable logistic regression with individual delivery as the unit of analysis.17 This logistic regression model included all covariates outlined above as well as an indicator variable for each of the 16 Health Services Delivery Areas in the province. We excluded the intercept term and centered all covariates on the study population mean proportion so that the risk-adjusted cesarean and assisted vaginal delivery rates would reflect an average mother in our study population.18 We considered Health Services Delivery Areas to have significantly higher or lower rates then the provincial mean when the 95% (confidence interval) for the risk-adjusted rate did not include the provincial rate for cesarean delivery and assisted vaginal delivery.17 We expected that, if cesarean deliveries were being performed for medically driven reasons, after adjusting for maternal characteristics and conditions there would be no regions with rates significantly different from the provincial average. Because of the large number of mothers missing valid BMI information, we ran the regressions with and without the variables representing mother's prepregnancy BMI and found no substantive difference in results; thus, we present the results of regressions excluding BMI. Since induction of labor may influence rates of operative delivery, but is not a medical indication for the procedure, we ran sensitivity analyses including and excluding this variable in the regressions. We found inclusion of induction of labor had a small effect on our adjusted cesarean delivery rates and thus included this variable in the final regression. All analyses were performed using Stata version 10.0 (StataCorp, College Station, TX).
Of the 153,458 deliveries in British Columbia between 2004 and 2007, there were 116,839 deliveries by mothers who had not had a previous cesarean delivery. Because of the large number of Health Services Delivery Areas in the province, we compared our study population's maternal characteristics and conditions across the five provincial health authorities—the larger health regions, which are composed of the 16 Health Services Delivery Areas (Table 1). The groups were comparable in mean weight gain during pregnancy, rates of multiple pregnancy, rates of breech or transverse presentation, and rates of preexisting diabetes. Rates of gestational diabetes and hypertension differed across the regions—mothers in the Northern Health Authority were less likely to have hypertension than mothers in the rest of province, and gestational diabetes rates were lower in both the Interior and Northern Health Authority. Age, BMI, and maternal smoking also differed markedly across the health authorities. Mothers from Vancouver Coastal Health were older, less likely to be overweight or obese, more likely to nulliparous, more likely to be diagnosed with hypertension, and less likely to smoke than mothers in the rest of the province.
Table 2 illustrates the total number of deliveries and the number and rates of primary cesarean deliveries, assisted vaginal deliveries, spontaneous vaginal deliveries, and inductions of labor. The provincial rate of primary cesarean delivery was 21.2 per 100 deliveries while the rate of assisted vaginal delivery was 14.2 per 100 deliveries. Crude primary cesarean and assisted vaginal delivery rates varied markedly across the Health Services Delivery Areas ranging from 16.1 to 27.5 per 100 deliveries and from 8.6 to 18.6 per 100 deliveries, respectively. The provincial rate of labor induction varied less than that of cesarean delivery and assisted vaginal delivery from 18.7 to 26.4 per 100 deliveries. Finally spontaneous vaginal delivery rates varied from 63.4 to 75.8 per 100 deliveries.
Table 3 shows crude primary cesarean delivery rates by indication for the surgical procedure across the 16 Health Services Delivery Areas. The most common indication for cesarean delivery was dystocia (which accounted for 30.0% of all cesarean delivery deliveries) followed by nonreassuring fetal heart rate (which accounted for 19.1% of all cesarean deliveries). While the most variable indication for cesarean delivery was declining a VBAC or maternal request for cesarean delivery, this indication represented only 2.0% of total cesarean deliveries in the province. Dystocia was the second most variable indication varying more than fivefold across the Health Services Delivery Areas (between 1.5 and 8.0 per 100 births). The least variable indication for cesarean delivery was breech presentation (which represented 16.3% of total cesarean deliveries in the province and varied between 2.1 and 4.0 per 100 births).
After controlling for maternal characteristics and conditions known to increase the likelihood of cesarean and assisted vaginal delivery, the rates of both still varied widely across regions. Table 4 presents crude and risk-adjusted rates for both modes of delivery. The risk-adjusted rates can be interpreted, as the region's rate had all mothers in that region, exhibited the characteristics of the average mother in the study population. Thus, regions whose risk-adjusted rates were higher than the crude rates performed cesarean deliveries more intensively than would have been predicted based on the observable characteristics and conditions of their mothers and vice versa. Risk-adjusted regional cesarean delivery rates revealed a nearly twofold difference, ranging from 14.7 to 27.6 per 100 deliveries. Adjusted assisted vaginal delivery rates varied by more than twofold, ranging from 6.5 to 15.3 per 100 deliveries. The adjusted cesarean delivery rates were significantly below the provincial rate in five regions and significantly above the provincial rate in three regions. The adjusted assisted vaginal delivery rates were significantly below the provincial average in ten regions and above the provincial average in two regions (although not significantly). The three regions where risk-adjusted cesarean delivery rates were significantly above the provincial rate all exhibited risk-adjusted assisted vaginal delivery rates significantly below the provincial rate, suggesting a potential trade-off between these modes of delivery in high cesarean delivery–regions.
This study did not attempt to determine the ideal cesarean delivery or assisted vaginal delivery rate; rather we examined the variation around the rate after controlling for medically indicated reasons for operative delivery. Our premise is that, regardless of the ideal rate, large-scale variation around the existing operative delivery rate can be reflective of uncertainty, inconsistency, or both, in obstetric medical care. We found that cesarean delivery and assisted vaginal delivery rates varied approximately twofold across British Columbia Health Services Delivery Areas and that much of this variation could not be explained by observable maternal characteristics and conditions that are known to increase the likelihood of both interventions during delivery. Our results also suggest that there appeared to be a trade-off between cesarean delivery and assisted vaginal delivery rates in regions with high risk-adjusted cesarean delivery rates, as all four regions with cesarean delivery rates significantly above the provincial rate also had assisted vaginal delivery rates significantly below the provincial rate. We also found that dystocia was the most common indication for cesarean delivery in British Columbia over our study period and that rates of cesarean delivery performed for this indication varied more than fivefold across the Health Services Delivery Areas.
Given that only 2.0% of cesarean deliveries were a result of maternal request or declining a VBAC, the wide variations in the rate of primary cesarean delivery could not be explained by regional variation in patient preferences. Thus, our results illustrate what we believe to be “unwarranted variation,” defined as variation that cannot be explained by the type or severity of illness or by patient preferences.13 This variation likely reflects differences in practitioners' approaches to medical decision-making, some of which may be explained by differences in the resources available to the practitioners. For example, smaller institutions often lack the resources required to respond to medical emergencies in the same manner as a tertiary care facility and would thus be more likely to recommend a cesarean delivery with a lower medical threshold than a larger institution. These hypotheses should be further investigated. Specifically, our findings suggest that differences in practitioners' response to dystocia should be investigated as a potentially important driver of regional variation.
There have been several American studies that have also reported significant variation in both cesarean delivery and assisted vaginal delivery rates. These studies have suggested that rates of operative delivery (both cesarean delivery and assisted vaginal delivery) in the United States are highly variable.11,19 Our results, although examining regional rather than temporal variation, are also consistent with a British Columbia report suggesting that maternal characteristics (such as increased age and prepregnancy weight) explained only a small amount of the temporal increase in cesarean delivery rates—a significant amount remained unexplained by changes in maternal characteristics over time.20 Both our results and the British Columbia report suggest that variations and increases in operative delivery rates have exceeded the variations in the prevalence of maternal characteristics and conditions that are known determinants of cesarean delivery.20 However, these results are inconsistent with a Canadian study of the Nova Scotia population, which found that temporal increases in the cesarean delivery rate could be explained by changes in maternal characteristics and conditions that resulted in obstetric practice changes over time.16
Our analysis is not without limitations, including those that are common when using large administrative datasets. The loss of some patients from our regression analyses due to missing BMI information is of potential concern. To minimize the likelihood that this missing data affected our results, we performed the regressions including BMI information and found the results were robust. We were also limited by our inability to control for the institution type in which the mother delivered. Previous research has shown that the odds of delivering by cesarean were three times higher in a tertiary hospital compared with a community hospital.21 Institution type is especially important given British Columbia is a large province with communities that are hundreds of miles from the nearest tertiary care facility. Those smaller community hospitals are likely to respond to a similar obstetric case slightly differently than a tertiary care facility. While the large variations we have reported are unlikely to have been driven by patient preference, patient choice could have potentially influenced the decision to perform assisted vaginal delivery compared with cesarean delivery. Finally, this study is simply the first step in using regional variation in cesarean and assisted vaginal delivery rates to shed light on differences in rates of operative delivery. While this study has determined that much of this variation cannot be explained by medical indications for cesarean delivery, further work should examine the role that nonmedical factors, such as socioeconomic status and the differences in provider mix and institutions, might play in driving regional variation.
In summary, we found significant regional variations in primary cesarean delivery and assisted vaginal delivery rates that could not be explained by regional differences in the maternal characteristics and conditions that are known determinants of operative delivery. Regardless of the actual ideal rates of operative delivery, our study results suggest that many regions are performing either too many or too few cesarean delivery deliveries. The apparent trade-off in many regions between assisted vaginal delivery and cesarean delivery suggests there may be variation in training with respect to operative delivery. There is most certainly variation in decision-making regarding when to perform assisted vaginal delivery compared with cesarean delivery. Specifically, when the fetal head is at a midpelvic or low midpelvic station, assisted vaginal delivery using instruments requires a significant degree of skill. We hypothesize that obstetricians with less practice performing assisted vaginal delivery in potentially complex situations may favor the use of the simpler cesarean delivery procedure—a hypothesis that should be further investigated especially given the difficulty that addressing such a skill deficit might pose.
Cesarean deliveries are important and life-saving interventions that have contributed significantly to the decrease in maternal and infant mortality rates. However, high cesarean delivery rates that are not fully medically driven are costly both to the health of mothers, and to the health care system as a whole. A cesarean delivery requires significant resources, including operating room time and substantially higher levels of health human resources than either an assisted vaginal delivery or a spontaneous vaginal delivery.8,9,22 Given the low VBAC rates, a cesarean delivery often commits the mother and her care providers to another cesarean delivery in the event of a subsequent pregnancy, extending the increased resource requirements and making the cumulative cost of a cesarean delivery (when considering subsequent births) significantly higher than spontaneous or assisted vaginal delivery.9 Given that our results suggest that dystocia is not only the most common indication for cesarean delivery, but also highly variable, we suggest that revising the current guidelines regarding the management of dystocia may be a good starting point on the road to decreasing unwarranted variation in cesarean delivery and assisted vaginal delivery rates.
2. Matthews TG, Crowley P, Chong A, McKenna P, McGarvey C, O'Regan M. Rising caesarean section rates: a cause for concern? BJOG 2003;110:346–9.
3. Foley ME, Alarab M, Daly L, Keane D, Macquillan K, O'Herlihy C. Term neonatal asphyxial seizures and peripartum deaths: lack of correlation with a rising cesarean delivery rate. Am J Obstet Gynecol 2005;192:102–8.
4. Daltveit AK, Tollanes MC, Pihlstrom H, Irgens LM. Cesarean delivery and subsequent pregnancies. Obstet Gynecol 2008;111:1327–34.
5. Wagner M. Choosing caesarean section. Lancet 2000;356:1677–80.
6. Liu S, Liston RM, Joseph KS, Heaman M, Sauve R, Kramer MS, et al. Maternal mortality and severe morbidity associated with low-risk planned cesarean delivery versus planned vaginal delivery at term. CMAJ 2007;176:455–60.
7. 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.
8. Allen VM, O'Connell CM, Farrell SA, Baskett TF. Economic implications of method of delivery. Am J Obstet Gynecol 2005;193:192–7.
9. Allen VM, O'Connell CM, Baskett TF. Cumulative economic implications of initial method of delivery. Obstet Gynecol 2006;108(3 Pt 1):549–55.
10. Society of Obstetricians and Gynaecologists of Canada. Rising C-section rates add risks during childbirth and place excess strain on the healthcare system, warn Canadian obstetricians. Ottawa (ON): SOGC Media Advisories; 2008.
11. Clark SL, Belfort MA, Hankins GD, Meyers JA, Houser FM. Variation in the rates of operative delivery in the united states. Am J Obstet Gynecol 2007;196:526.e1–5.
12. Jha AK, Li Z, Orav EJ, Epstein AM. Care in U.S. hospitals–the Hospital Quality Alliance Program. N Engl J Med 2005;353:265–74.
13. Wennberg JE. Unwarranted variations in healthcare delivery: Implications for academic medical centres. Br Med J 2002;325:961–4.
14. British Columbia reproductive care program. British columbia perinatal database registry overview. Vancouver (BC): British Columbia Perinatal Health Program; 2003.
15. Roberts RG, Deutchman M, King VJ, Fryer GE, Miyoshi TJ. Changing policies on vaginal birth after cesarean: impact on access. Birth 2007;34:316–22.
16. 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.
17. Thompson LA, Goodman DC, Chang CH, Stukel TA. Regional variation in rates of low birth weight. Pediatrics 2005;116:1114–21.
18. Vittinghoff E. Regression methods in biostatistics: Linear, logistic, survival, and repeated measures models. New York (NY): Springer; 2005.
19. Baicker K, Buckles KS, Chandra A. Geographic variation in the appropriate use of cesarean delivery. Health Aff 2006;25:355–67.
20. Caesarean birth task force report. Vancouver (BC): British Columbia Perinatal Health Program; 2008.
21. Janssen PA, Klein MC, Soolsma JH. Differences in institutional cesarean delivery rates-the role of pain management. J Fam Pract 2001;50:217–23.
22. Petrou S, Glazener C. The economic costs of alternative modes of delivery during the first two months postpartum: results from a Scottish observational study. BJOG 2002;109:214–7.
Figure. No caption available.