The labor-adjusted cesarean section rate--a more informative method than the cesarean section "rate" for assessing a practitioner's labor and delivery skills

Risk adjustment for inter-hospital comparison of primary cesarean section rates: need, validity and parsimony

BACKGROUND

Cesarean section rates is often used as an indicator of quality of care in maternity hospitals. The assumption is that lower rates reflect in developed countries more appropriate clinical practice and general better performances. Hospitals are thus often ranked on the basis of caesarean section rates. The aim of this study is to assess whether the adjustment for clinical and sociodemographic variables of the mother and the fetus is necessary for inter-hospital comparisons of cesarean section (c-section) rates and to assess whether a risk adjustment model based on a limited number of variables could be identified and used.

METHODS

Discharge abstracts of labouring women without prior cesarean were linked with abstracts of newborns discharged from 29 hospitals of the Emilia-Romagna Region (Italy) from 2003 to 2004. Adjusted ORs of cesarean by hospital were estimated by using two logistic regression models: 1) a full model including the potential confounders selected by a backward procedure; 2) a parsimonious model including only actual confounders identified by the "change-in-estimate" procedure. Hospital rankings, based on ORs were examined.

RESULTS

24 risk factors for c-section were included in the full model and 7 (marital status, maternal age, infant weight, fetopelvic disproportion, eclampsia or pre-eclampsia, placenta previa/abruptio placentae, malposition/malpresentation) in the parsimonious model. Hospital ranking using the adjusted ORs from both models was different from that obtained using the crude ORs. The correlation between the rankings of the two models was 0.92. The crude ORs were smaller than ORs adjusted by both models, with the parsimonious ones producing more precise estimates.

CONCLUSION

Risk adjustment is necessary to compare hospital c-section rates, it shows differences in rankings and highlights inappropriateness of some hospitals. By adjusting for only actual confounders valid and more precise estimates could be obtained.

Effects of the DRG-based prospective payment system operated by the voluntarily participating providers on the cesarean section rates in Korea

This study explored the effects of the diagnosis-related group (DRG)-based prospective payment system (PPS) operated by voluntarily participating organizations on the cesarean section (CS) rates, and analyzed whether the participating health care organizations had similar CS rates despite the varied participation periods. The study sample included delivery claims data from the Korean national health insurance program for the year 2003. Risk factors were identified and used in the adjustment model to distinguish the main reason for CS. Their risk-adjusted CS rates were compared by the reimbursement methods, and the organizations' internal and external environments were controlled. The final risk-adjustment model for the CS rates meets the criteria for an effective model. There were no significant differences of CS rates between providers in the DRG and fee-for-service system after controlling for organizational variables. The CS rates did not vary significantly depending on the providers' DRG participation periods. The results provide evidence that the DRG payment system operated by volunteering health care organizations had no impact on the CS rates, which can lower the quality of care. Although the providers joined the DRG system in different years, there were no differences in the CS rates among the DRG providers. These results support the future expansion of the DRG-based PPS plan to all health care services in Korea.

How much variation in CS rates can be explained by case mix differences?

OBJECTIVE

To quantify the amount of variation in caesarean section (CS) rates between maternity units explained by case mix differences.

DESIGN

Cross-sectional study.

SETTING

All 216 maternity units in England and Wales.

POPULATION

Women giving birth at these maternity units between May and July 2000.

METHODS

Logistic regression models were developed to investigate the relationship between case mix characteristics, and odds of (i) CS before labour, (ii) CS in labour. Using these results, overall CS rates standardised for case mix were calculated for each maternity unit. Random-effects meta-analysis was used to examine heterogeneity between maternity units.

MAIN OUTCOME MEASURES

CS before labour and CS during labour.

RESULTS

Adjustment for case mix differences between maternity units explained 34% of the variance in CS rates. Odds of CS (before and in labour) increased with maternal age. Women from ethnic minority groups had lower odds of CS before labour, and increased odds of CS in labour. Women with a previous vaginal delivery had lower odds of CS, although the magnitude of this for CS before and in labour is markedly different.

CONCLUSIONS

Case mix adjustment is important to enable understanding of the factors that influence the CS rate. These include organisational and staffing levels as well as women's preferences for childbirth and clinician's attitudes. An understanding of how these factors influence the CS rate is essential for evaluation of quality and appropriateness of obstetric care provided to women.

Physician contribution to a cesarean delivery risk model

OBJECTIVE

The purpose of this study was to assess the contribution of physician management to the probability of cesarean delivery.

STUDY DESIGN

A prospective cohort study was performed of all live births who weighed > or =500 g for a 2-year period (1999-2000) at a large metropolitan hospital (n = 10,027 births). Factors that were associated significantly with cesarean delivery at one time excluded cases in which cesarean delivery was a necessary or probable outcome. In the planned vaginal delivery sample (n = 7940 births), a risk-adjusted logistic regression model was used to assess the prediction of cesarean delivery. To test for the effect of physician-management physician, we used the subset of physicians with > or =45 deliveries in the 2-year time period (n = 6563 deliveries).

RESULTS

When physician-management physician data were added to the use of forward stepwise regression, entry order was abnormal position, nulliparity, birth weight of >4000 g, and physician. The model's predictive ability improved from 43.8% to 50.2%.

CONCLUSION

Physician management adds a significant independent effect to the cesarean delivery risk model.

Using administrative data to identify indications for elective primary cesarean delivery

OBJECTIVE

To develop a methodology to identify indications and normative rates for elective primary cesarean delivery using administrative data.

DATA SOURCES/STUDY SETTING

All delivery discharges in 1995, as reported to the California Office of Statewide Health Planning and Development (secondary data).

STUDY DESIGN

Retrospective population based study.

DATA COLLECTION/EXTRACTION

Data were entered into a recursive partitioning algorithm to develop a hierarchy of conditions by which patients with multiple conditions could be sorted with respect to the binary outcome of labor or elective primary cesarean without labor. This hierarchy was examined for its clinical consistency, validated on a second sample, and compared with results obtained from logistic regression.

PRINCIPAL FINDINGS

Four percent (19,664) of delivery discharges in 1995 underwent elective primary cesarean. Twelve clinical conditions contributed to the hierarchy, and accounted for 92.9 percent of all women experiencing elective primary cesarean delivery. The remaining 7.1 percent of the elective primary cesarean cases were classified as "unspecified."

CONCLUSIONS

A standardized methodology (utilizing recursive partitioning algorithms) for assigning indications for elective primary cesarean is presented. This methodology relies on administrative data, classifies women with complex comorbidity patterns into clinically relevant subpopulations, and can be used to establish normative rates for benchmarking specific indications for cesarean delivery.

Caesarean section rate for maternal indication in sub-Saharan Africa: a systematic review

INTRODUCTION

Rates of caesarean sections in more-developed countries have been rising since 1970, and vary greatly between less-developed countries. Present estimates, based on data from more-developed countries need to be validated with data from less-developed countries. We estimated the need for caesarean section for maternal indication in a population of pregnant women in west Africa (MOMA survey).

METHODS

The expected caesarean section rate was calculated from the rate of obstetric risk in the MOMA population, and rates of caesarean section in published work.

FINDINGS

Three-quarters of women from hospitals of sub-Saharan Africa were delivered by caesarean section for maternal reasons. Such intervention was needed for six main reasons, protracted labour, abruptio placentae, previous caesarean section, eclampsia, placenta praevia, and malpresentation. Although the observed rate of caesarean section in west African women is 1.3%, our results, combined with those of published work suggest a range of 3.6-6.5% (median, 5.4%).

INTERPRETATION

Our method might not be strictly accurate, but it is simple and provides informative findings that can help policy makers and health planners in sub-Saharan Africa to design and follow up programmes to reach the optimum caesarean section rate. Moreover, application of this method to hospital data could improve practitioners' assessments in these countries.

Variation in elective primary cesarean delivery by patient and hospital factors

OBJECTIVE

Our purpose was to describe variation in elective primary cesarean rates by nonclinical factors.

STUDY DESIGN

With use of California discharge data and American Hospital Association data for 1995, patients were classified into 13 mutually exclusive categories for elective primary cesarean delivery. With use of recursive partitioning algorithms, women in each category were then studied to determine whether nonclinical factors were associated with elective primary cesarean delivery.

RESULTS

A total of 463,196 women were delivered at 288 hospitals, and the elective primary cesarean delivery rate was 4.25% (19,664/463,196). Risk for elective primary cesarean delivery varied by clinical condition. The most discriminant risk factors were hospital type (malpresentation, multiple gestation, macrosomia, other hypertension), maternal age (antepartum bleeding, uterine scar, soft tissue disorder, preterm, unspecified), and teaching status (herpes, severe hypertension, unengaged head).

CONCLUSION

This article presents methods that use administrative data to isolate and monitor the impact of nonclinical factors on the use of elective primary cesarean.

Decision-case mix model for analyzing variation in cesarean rates

This article contributes a decision-case mix model for analyzing variation in c-section rates. Like recent contributions to the literature, the model systematically takes into account the effect of case mix. Going beyond past research, the model highlights differences in physician decision making in response to obstetric factors. Distinguishing the effects of physician decision making and case mix is important in understanding why c-section rates vary and in developing programs to effect change in physician behavior. The model was applied to a sample of deliveries at a hospital where physicians exhibited considerable variation in their c-section rates. Comparing groups with a low versus high rate, the authors' general conclusion is that the difference in physician decision tendencies (to perform a c-section), in response to specific obstetric factors, is at least as important as case mix in explaining variation in c-section rates. The exact effects of decision making versus case mix depend on how the model application defines the obstetric condition of interest and on the weighting of deliveries by their estimated "risk of Cesarean." The general conclusion is supported by an additional analysis that uses the model's elements to predict individual physicians' annual c-section rates.

Adjustment for case mix in comparisons of cesarean delivery rates: university versus community hospitals in Vermont

OBJECTIVE

Our objective was to determine whether case mix model adjustment would help to explain differences in cesarean delivery rates between community and university hospitals. We also wished to define a patient population in which the cesarean delivery rate would be more reflective of individual practice patterns than of obstetric or medical risk.

STUDY DESIGN

Established risk factors for cesarean delivery were identified by retrospective chart review at two community hospitals (designated A and B) and a university hospital. Each delivery was assigned exclusively to 1 of 6 risk categories: (1) multiple gestation, (2) fetal malpresentation, (3) delivery at <36 weeks' gestation, (4) not suitable for trial of labor, and (5) term delivery (> or =36 weeks' gestation) with medical complications, and (6) term delivery (> or =36 weeks' gestation) without medical complications. Parity and history of cesarean delivery further subdivided these categories into a total of 18 unique subgroups. Case mix was defined as the distribution of patients into each subgroup. Patients assigned to the categories of multiple gestation, fetal malpresentation, delivery at <36 weeks' gestation, and not eligible for trial of labor were considered to compose the group at high risk for cesarean delivery. The remaining patients composed the group at low risk for cesarean delivery. Observed cesarean delivery rates were calculated for each cell of the case mix grid within individual hospitals. Total, primary, and repeat cesarean delivery rates were determined for each hospital. The cesarean delivery rates for the low-risk populations were calculated. Data were evaluated both by chi(2) test and by direct standardization analysis with the university hospital case mix used as the standard population.

RESULTS

A total of 5705 delivery reports were reviewed (university hospital, n = 4538; hospital A, n = 531; hospital B, n = 636). The cesarean delivery rates were significantly different between hospitals (university hospital, 16. 9%; hospital A, 13.6%; hospital B, 12.0%; P =.002). The distributions of patients in the high-risk group were also significantly different between hospitals (university hospital, 16. 8%; hospital A, 5.8%; hospital B, 8.8%; P = .001). The percentage of medically complicated cases in the low risk for cesarean group was significantly higher at the university hospital (university hospital, 16.9%; hospital A, 8.8%; hospital B, 9.8%; P =.001). However, no statistical differences were detected between hospitals in either the observed cesarean delivery rates or the standardized rates for the low-risk groups.

CONCLUSION

The case mix model provides a more accurate method of comparing cesarean delivery rates between community and university hospitals. The low-risk group of patients discriminated in this model represents a population in which the cesarean delivery rate may be more reflective of individual practice patterns than of maternal or fetal risks.

Initial development of a system-wide maternal-fetal outcomes assessment program

OBJECTIVE

This study was undertaken to develop a comprehensive risk-assessment approach capable of evaluating maternal and fetal outcomes.

STUDY DESIGN

Data from 10,984 women and 11,066 infants delivered at 79 military treatment facilities in the United States from 1995 to 1997 were used to develop two individual but complementary risk-adjustment models for maternal and, separately, fetal outcomes. A range of maternal and delivery-related risk variables and clinically important outcomes were identified by expert opinion and selected and weighted with ordinal logistic regression analysis. Receiver operating characteristic curves for the maternal and fetal models were determined. Variation across the facilities in risk-adjusted performance was also evaluated.

RESULTS

Risk factors and poor outcomes were rare for both mothers and infants, with 96.9% of infants and 97.7% of mothers having good or excellent outcomes (0.7% mortality and 0.01% mortality, respectively). Despite the low frequency of poor outcomes both models performed well, with receiver operating characteristic curves of 0.75 for maternal outcomes and 0.78 for infant outcomes. When the models were applied to the military treatment facilities, there were significant differences among facilities in risk-adjusted outcomes. Twenty-four of the facilities in the study (30%) had outcomes odds ratios that were significantly >1 or significantly <1 (P <.05). There did not appear to be any relationship between the performance of a military treatment facility for maternal outcome and that for infant outcome.

CONCLUSION

Complementary risk models for maternal and infant outcomes were developed that had satisfactory discriminatory power across a variety of facilities within a large health system. With further development and refinement this approach holds promise of being able to detect variations in risk-adjusted performance that could be used to identify best practices. The results might also be used to help coordinate and improve the quality of care for the entire conception-to-delivery process.

Influence of maternal-fetal medicine subspecialization on the frequency of trial of labor in term pregnancies with breech presentation

OBJECTIVE

To investigate the role of subspecialization in maternal-fetal medicine (MFM) on the frequency of a trial of labor in term pregnancies with breech presentation.

METHODS

We conducted a retrospective study of 332 singleton pregnancies > or =37 weeks with nonfootling breech presentation that delivered over a 6-year period (1994-1998) at a university-based, tertiary care hospital. Patients were divided into two groups based on whether the delivery was attended by an MFM or non-MFM obstetrician-gynecologist. Demographic and clinical data were compared between groups and outcome variables included whether the patient had an attempt at vaginal delivery, cesarean delivery after a labor attempt, or vaginal breech delivery.

RESULTS

The frequency of labor attempt (OR 1.4, 95% CI 0.9-2.3), vaginal breech success rate (OR 0.6, 95% CI 0.3-1.5), and overall cesarean rates (OR 0.9, 95% CI 0.5-1.7) were similar between groups. Using discriminant function analysis, only nulliparity (R2 = 1.6%, F = 6.0, P = 0.005) and birthweight (R2 = 2.0% F = 6.4, P = 0.01) were associated with trial of vaginal delivery.

CONCLUSIONS

Subspecialization in MFM had no impact on the frequency of trial of labor in the term pregnancy with a breech presentation.

Cesarean delivery risk adjustment for regional interhospital comparisons

OBJECTIVE

The aim of this study was to determine the effect of adjustment for patient population mix on observed, expected, and standardized cesarean delivery rates in regional hospitals.

STUDY DESIGN

Multiple logistic regression was applied to a large regional perinatal database comprising 16 hospitals. Variables significantly associated with cesarean delivery were used to calculate cesarean delivery probabilities for individual patients. Probabilities were summed across hospitals to derive expected hospital cesarean delivery rates. A standardized rate for each hospital was then calculated by dividing the observed rate by the expected rate and multiplying by the regional rate.

RESULTS

The regional cesarean delivery rate was 21.9% for 6798 women. Observed hospital rates varied from 17.1% to 39.2%. Twenty-two variables were associated with cesarean delivery. Expected cesarean delivery rates ranged from 18.1% to 26.0%. Among the 5 hospitals with the lowest observed cesarean delivery rates only 2 had rates significantly lower than those of the rest of the region, and only 1 of those 2 rates remained significantly lower after adjustment. One other hospital that had an adjusted rate significantly lower than the crude rate had not appeared statistically different from the rest of the region before standardization. Among the 5 hospitals with the highest cesarean delivery rates, 4 had rates significantly higher than the rest of the region, and 3 of them had significantly higher observed rates than expected rates.

CONCLUSIONS

Compared with using observed (crude) cesarean delivery rates, adjustment for differences in patient risk factor mix facilitates more accurate comparison of cesarean delivery rates among hospitals within a region.

Cesarean delivery rates in the United States. The 1990s

The increase in CS rates in the United States in the 1970s and 1980s and the gradual decrease in the 1990s have been the focus of considerable attention because of the increased maternal morbidity and cost associated with the procedure without apparent impact on infant mortality. Focused efforts to reduce CS have resulted in a modest decrease the rate of primary CS and a marked increase in VBAC. Considerable variation in CS rates exists among regions in the United States and among states within those regions. The states with the higher CS rates are clustered in the South and Northeast regions of the United States, whereas rates tend to be lower in the West and Midwest. This variation cannot be explained by standard demographic risk factors and is likely related to local culture and mode of practice. Patient case mix should also be taken into account when comparing CS rates. Accounting for differences risk may help highlight differences in mode of practice and thus identify opportunities for improvement. Several reports from hospitals and communities of education and peer review programs have resulted in a significant reduction in their CS rates without increasing perinatal or maternal morbidity and mortality. A common theme in these reports of successful strategies to decrease the CS rate safely is the importance of physician motivation to make a change.

Institutional influences on the primary cesarean section rate in Utah, 1992 to 1995

OBJECTIVES

Our purpose was to evaluate institutional and organizational influences on cesarean section rates in Utah and to adjust such rates for differences in patient acuity.

STUDY DESIGN

Data on cesarean section rates were derived from the Utah Hospital Discharge Database and adjusted for patient acuity by correcting raw cesarean rates for those patients undergoing cesarean section meeting regional gestational age transport criteria.

RESULTS

When analyzed by means of 1-way analysis of variance, the following factors had a significant negative correlation (P < .05) with cesarean section rate: presence of a newborn intensive care unit and maternal-fetal medicine subspecialists, presence on the medical staff of obstetrician-gynecologist(s) as opposed to family physicians only, delivery volume >1500/y, urban location, and 24-hour in-house anesthesiology. When cesarean rates were corrected for acuity, facilities with maternal-fetal medicine specialists and a newborn intensive care unit had significantly lower rates (P < .001) and more uniform rates than otherwise similar institutions.

CONCLUSIONS

More medically sophisticated physicians and institutions have lower cesarean rates when patient acuity is taken into account.