Objective
The purpose of this study was to examine the relationship between delivery volume and maternal complications.
Study Design
We used administrative data to identify women who had been admitted for childbirth in 2006. Hospitals were stratified into deciles that were based on delivery volume. We compared composite complication rates across deciles.
Results
We evaluated 1,683,754 childbirths in 1045 hospitals. Decile 1 and 2 hospitals had significantly higher rates of composite complications than decile 10 (11.8% and 10.1% vs 8.5%, respectively; P < .0001). Decile 9 and 10 hospitals had modestly higher composite complications as compared with decile 6 (8.8% and 8.5% vs 7.6%, respectively; P < .0001). Sixty percent of decile 1 and 2 hospitals were located within 25 miles of the nearest greater volume hospital.
Conclusion
Women who deliver at very low-volume hospitals have higher complication rates, as do women who deliver at exceedingly high-volume hospitals. Most women who deliver in extremely low-volume hospitals have a higher volume hospital located within 25 miles.
More than 4 million women give birth annually in the United States, which makes childbirth the single most common reason for hospitalization among young women. Childbirth in the United States generally is safe with a major complication rate (eg, hemorrhage or infection) of <10%. At the same time, there is growing appreciation that the variation in hospital outcomes that have been observed in many medical and surgical diagnoses may also exist for childbirth.
For Editors’ Commentary, see Contents
See related editorial, page 1
More than 2 decades of research has documented a relationship between higher hospital volume and improved clinical outcomes for medical and surgical diagnoses, but data in the area of maternal childbirth outcomes are sorely lacking. The paucity of empiric studies of the volume-outcome relationship for maternal complications is striking, given the clinical volume and economic impact of childbirth in the United States. In addition, methodologic limitations of the few published studies that have related to a volume-outcome relationship for the maternal outcomes of childbirth make interpretation of the conflicting findings difficult.
Our objective was to examine rigorously the relationship between hospital volume and maternal childbirth outcomes. Specifically, we set out to examine the association between hospital childbirth volume and important maternal complications (eg, hemorrhage, infection, death) and whether the volume-outcome relationship might differ for vaginal and cesarean deliveries.
Materials and Methods
We used a 100% sample of State Inpatient Data (SID) for year 2006 from 11 states (Arizona, California, Florida, Iowa, Massachusetts, Maryland, North Carolina, New York, New Jersey, Washington, and Wisconsin) to identify all patients who were hospitalized with childbirth ( International Classification of Diseases , 9th Clinical Modification [ICD-9-CM] code 650 or 640x to 676.9x [n = 1,683,754]). Maternal childbirth admissions were stratified into cesarean deliveries (ICD-9 procedure code 74) or normal spontaneous vaginal deliveries (ICD-9 diagnosis codes 640.x to 676.9x and the absence of a code for cesarean delivery).
We excluded patients who delivered after transfer from another acute care hospital because transfer patients are more complex than patients who are admitted through other routes and because administrative data do not adequately capture this excess complexity that leads to potentially biased results. All other cases of childbirth (spontaneous vaginal deliveries, cesarean deliveries, forceps, and vacuum) were included in the analysis of “all deliveries” ( Figure 1 ). Our analysis of spontaneous vaginal deliveries excluded instrument deliveries with forceps or vacuums because the use of such devices suggests a more complicated delivery and may introduce unwanted heterogeneity. Application of these criteria left us with a cohort of what we would describe as routine childbirth admissions without obvious evidence of complicating factors.
The SID databases that were used in this study were developed by the Agency for Healthcare Research and Quality as part of the health care utilization project in partnership with individual states ( http://hcupnet.ahrq.gov/ ). We deliberately acquired SID data from the 11 states that were included in this study because they represented all regions of the United States, because they included a disproportionate percentage of the US population, and because they covered a mix of urban and rural regions. SID data include many elements that are included on the Uniform Billing claim form (UB-92) hospital discharge abstract and have been used extensively in previous health services research, including previous obstetrics studies. Key data elements include patient demographics, admitting hospital, primary and secondary diagnoses and procedures (as captured by ICD-9-CM codes), the diagnosis related group, admission source (eg, emergency department, transfer from another hospital), admission and discharge dates, patient’s primary insurance (categorized as Medicare, private insurance, Medicaid, self-pay, other), type of insurance (fee-for-service or health maintenance organization), and disposition at the time of hospital discharge (eg, transfer to another acute care hospital, deceased).
We calculated 3 separate measures of childbirth volume for each hospital by summing the total number of deliveries that were performed during 2006: (1) total childbirth volume, (2) spontaneous vaginal delivery volume, and (3) cesarean delivery volume. Hospitals were then stratified into deciles of volume for each of the 3 delivery categories; thus, a hospital could be in the highest (tenth) decile of volume for total childbirth but the eighth decile for vaginal deliveries.
We identified 6 key adverse outcomes of childbirth that have been evaluated in previous studies using administrative data that included hemorrhage, severe perineal lacerations (3rd- or 4th-degree lacerations), operative complications, infection, thrombotic complications, and death. Outcomes of interest for vaginal deliveries included all of the outcomes described, with the exception of operative complications that are not relevant to vaginal delivery. Outcomes of interest for cesarean deliveries included all outcomes, except for severe perineal lacerations that are not relevant to cesarean deliveries. From an analytic standpoint, our primary outcome was a composite measure that represented the occurrence of ≥1 adverse outcomes in a given patient. We identified comorbid illnesses using the method developed by Elixhauser et al and supplemented this by high-risk obstetrical conditions that have been identified previously.
Statistical analysis
We used bivariate methods that included the t -test and Cochran-Mantel-Haenszel statistics to compare the demographic characteristics (eg, age, race) of patients across hospital volume deciles. We used similar methods to compare insurance coverage and the incidence of comorbid illnesses across deciles of hospital volume. All analyses were conducted separately for spontaneous vaginal deliveries, cesarean section deliveries, and all deliveries in aggregate. We used similar methods to compare the unadjusted incidence of in-hospital maternal complications across hospital volume deciles.
Finally, we used a series of logistic regression models to evaluate the association between rates of childbirth complications and hospital obstetrics volume after adjustment for differences in patient demographics and comorbidity. We used the patient as the unit of analysis; volume was measured at the hospital level. The standard errors, 95% confidence intervals, and associated significance levels for adjusted odds ratios based on logistic regression accounted for clustering with the use of hospital random-effect models. For purposes of these analyses, the outcome (dependent variable) was an indicator variable that represented the occurrence of the composite outcome. The dependent variables of interest were a series of indicator variables that represented the decile of hospital volume, with hospitals in the highest volume decile serving as the reference category. The models included 19 covariates that included patient age, race, payor, and a number of important comorbid illness. Separate analyses were performed for (1) all childbirths in aggregate, (2) spontaneous vaginal deliveries only, and (3) cesarean deliveries only. We applied 3 separate models to each patient cohort: (1) unadjusted, (2) adjustment for patient demographics alone, and (3) adjustment for patient demographics plus comorbidities.
Because we hypothesized that more complex cases would be referred selectively to higher volume hospitals, we conducted a number of sensitivity analyses. Specifically, we stratified both vaginal deliveries and cesarean deliveries into high-risk and low-risk cohorts. High-risk patients were defined as those with any of the following conditions that typically are considered a cause for concern among obstetricians: advanced age, asthma, cerebral hemorrhage, hypertensive disorders, diabetes mellitus, obesity, chorioamnionitis, congenital heart disease, liver anomalies, renal anomalies, thyroid disease, mental disorder, multiple gestation, preterm gestation, pulmonary embolism, and uterine rupture. The low-risk cohort included patients without any of these conditions. We also replicated our analyses to examine alternative methods for categorizing hospital volume (eg, quintiles, quartiles) and defining high- and low-volume hospitals.
All statistical analyses were performed with SAS software (version 9.2; SAS Institute Inc., Cary, NC). The study was approved by the Institutional Review Board of the University of Iowa.
Results
We identified a total of 1,683,754 childbirths. After excluding transfer cases (n = 4945), our final cohort of spontaneous vaginal deliveries, forceps deliveries, vacuum extractions, and cesarean deliveries included 1,678,809 admissions to 1045 hospitals. After the exclusion of forceps and vacuum-assisted deliveries (n = 94,188), there were a total of 1,047,848 spontaneous vaginal deliveries in 1011 hospitals (34 hospitals delivered only by cesarean section) and 536,773 cesarean deliveries in 1030 hospitals (15 hospitals delivered only by vaginal route). For hospitals that performed both vaginal and cesarean deliveries, the cesarean delivery rate ranged from 13.0–96.7% across 1030 hospitals (mean, 31.3%; median, 30.3%). The mean hospital childbirth volume was 1606.5 (interquartile range, 442.0–2299.0); the mean vaginal delivery volume was 1036.4 (interquartile range, 308.0–1472.0), and the mean cesarean delivery volume was 521.1 (interquartile range, 135–754).
The characteristics of patients with spontaneous vaginal delivery that were stratified by hospital volume deciles are displayed in Table 1 . Women who delivered in lower volume hospitals tended to younger, were more likely to be white, and were more likely to be categorized as self-pay when compared with women who delivered at higher volume hospitals. Women who underwent vaginal delivery in lower volume hospitals tended to have fewer comorbid conditions such as advanced maternal age, hypertension, and diabetes mellitus ( Table 1 ). Conversely, women who had vaginal deliveries in high-volume hospitals were more likely to be older and Hispanic or black and to have more comorbid illness. Results were similar for cesarean deliveries ( Table 2 ), with low-volume hospitals treating a higher proportion of uninsured younger white women, although high-volume hospitals treated more women with advanced age and comorbid illness.
| Vaginal deliveries | Decile 1 | Decile 2 | Decile 3 | Decile 4 | Decile 5 | Decile 6 | Decile 7 | Decile 8 | Decile 9 | Decile 10 |
|---|---|---|---|---|---|---|---|---|---|---|
| General characteristics | ||||||||||
| Hospitals, n | 105 | 104 | 104 | 105 | 104 | 105 | 105 | 104 | 105 | 104 |
| Range of delivery volume, n | 1–99 | 100–307 | 308–551 | 552–788 | 789–1133 | 1134–1567 | 1568–2048 | 2049–2675 | 2676–3724 | 3725–12,845 |
| Specific characteristics | ||||||||||
| Patients, n | 1930 | 13,299 | 28,507 | 48,108 | 63,105 | 88,305 | 117,133 | 150,944 | 210,723 | 328,794 |
| Delivery volume, n a | 44.3 ± 33.2 | 165.6 ± 39.4 | 309.8 ± 48.6 | 464.3 ± 43.0 | 643.0 ± 54.2 | 868.6 ± 81.8 | 1139.8 ± 78.9 | 1489.0 ± 130.7 | 2010.8 ± 186.1 | 3231.4 ± 919.9 |
| High-risk volume, n a | 9.1 ± 6.6 | 33.7 ± 16.1 | 67.1 ± 29.5 | 107.7 ± 40.2 | 155.0 ± 57.7 | 217.6 ± 77.1 | 306.4 ± 120.7 | 400.2 ± 144.1 | 541.0 ± 194.2 | 941.0 ± 402.0 |
| Low-risk volume, n a | 38.7 ± 27.6 | 131.9 ± 33.7 | 242.7 ± 43.3 | 356.6 ± 49.3 | 488.0 ± 68.6 | 651.1 ± 88.6 | 833.4 ± 135.2 | 1088.8 ± 153.8 | 1469.8 ± 227.3 | 2290.3 ± 683.7 |
| Age, y a | 26.0 ± 5.6 | 25.9 ± 5.7 | 26.0 ± 5.9 | 26.4 ± 6.0 | 26.8 ± 6.1 | 27.2 ± 6.1 | 27.4 ± 6.2 | 27.2 ± 6.1 | 27.3 ± 6.1 | 27.9 ± 6.2 |
| Race, n (%) | ||||||||||
| White | 1445 (74.9) | 8807 (66.2) | 15,707 (55.1) | 25,333 (56.2) | 28,628 (45.4) | 41,419 (46.9) | 44,652 (38.1) | 60,716 (40.2) | 77,926 (37.0) | 119,403 (36.3) |
| Hispanic | 63 (3.3) | 728 (5.5) | 3473 (12.2) | 5627 (12.5) | 14,030 (22.2) | 19,767 (22.4) | 33,203 (28.4) | 41,500 (27.5) | 64,437 (30.6) | 96,110 (29.2) |
| Black | 59 (3.1) | 372 (2.8) | 1237 (4.3) | 2165 (4.8) | 5085 (8.1) | 10,031 (11.4) | 14,143 (12.1) | 21,137 (14.0) | 18,876 (9.0) | 36,585 (11.1) |
| Other | 363 (18.8) | 3392 (25.5) | 8090 (28.4) | 11,983 (26.6) | 15,362 (24.3) | 17,088 (19.4) | 25,135 (21.5) | 27,591 (18.3) | 49,484 (23.5) | 76,696 (23.3) |
| Payor, n (%) | ||||||||||
| Medicaid | 858 (44.5) | 6694 (50.3) | 14,324 (50.3) | 21,108 (46.7) | 28,514 (45.2) | 36,643 (41.5) | 54,524 (46.6) | 66,010 (43.7) | 93,202 (44.2) | 142,979 (43.5) |
| Private | 862 (44.7) | 5,734 (43.1) | 11,682 (41.0) | 21,482 (47.6) | 28,885 (45.8) | 45,110 (51.1) | 54,512 (46.5) | 73,341 (48.6) | 105,493 (50.1) | 170,092 (51.7) |
| Medicare | 8 (0.4) | 36 (0.3) | 200 (0.7) | 174 (0.4) | 255 (0.4) | 340 (0.4) | 338 (0.3) | 549 (0.4) | 770 (0.4) | 643 (0.2) |
| Self-pay | 164 (8.5) | 531 (4.0) | 1077 (3.8) | 1377 (3.1) | 3682 (5.8) | 3740 (4.2) | 5575 (4.8) | 7995 (5.3) | 6772 (3.2) | 9650 (2.9) |
| Comorbidities, n | ||||||||||
| Advanced maternal age, n (%) | 166 (8.6) | 1177 (8.9) | 2710 (9.5) | 4905 (10.9) | 7781 (12.3) | 11,935 (13.5) | 16,502 (14.1) | 20,323 (13.5) | 28,322 (13.4) | 52,611 (16.0) |
| Hypertension disorder, n (%) | 33 (1.7) | 277 (1.3) | 661 (2.3) | 885 (2.0) | 1363 (2.2) | 1867 (2.1) | 2712 (2.3) | 3455 (2.3) | 4797 (2.3) | 9210 (2.8) |
| Diabetes mellitus, n (%) | 22 (1.1) | 222 (1.7) | 500 (1.8) | 891 (2.0) | 1436 (2.3) | 2016 (2.3) | 3114 (2.7) | 4043 (2.7) | 5814 (2.8) | 10,126 (3.1) |
| Obesity, n (%) | 10 (0.5) | 94 (0.7) | 194 (0.7) | 358 (0.8) | 476 (0.8) | 788 (0.9) | 1287 (1.1) | 1565 (1.0) | 3068 (1.5) | 2758 (0.8) |
| Multiple gestation, n (%) | 2 (0.1) | 11 (0.1) | 21 (0.01) | 59 (0.1) | 72 (0.1) | 161 (0.2) | 222 (0.2) | 306 (0.2) | 494 (0.2) | 1032 (0.3) |
| Preterm gestation, n (%) | 18 (0.9) | 185 (1.4) | 452 (1.6) | 727 (1.6) | 1209 (1.9) | 1801 (2.0) | 2674 (2.3) | 3879 (2.6) | 5783 (2.7) | 10,699 (3.3) |
| Cesarean deliveries | Decile 1 | Decile 2 | Decile 3 | Decile 4 | Decile 5 | Decile 6 | Decile 7 | Decile 8 | Decile 9 | Decile 10 |
|---|---|---|---|---|---|---|---|---|---|---|
| General characteristics | ||||||||||
| Hospitals, n | 105 | 104 | 104 | 105 | 104 | 105 | 105 | 104 | 105 | 104 |
| Range of delivery volume, n | 1–99 | 100–307 | 308–551 | 552–788 | 789–1133 | 1134–1567 | 1568–2048 | 2049–2675 | 2676–3724 | 3725–12,845 |
| Specific characteristics | ||||||||||
| Patients, n | 922 | 6006 | 14,175 | 20,779 | 30,229 | 45,435 | 60,659 | 79,007 | 102,481 | 177,080 |
| Volume, n a | 13.2 ± 11.8 | 64.1 ± 17.3 | 135.5 ± 21.8 | 208.1 ± 19.9 | 291.1 ± 33.9 | 422.1 ± 45.3 | 575.2 ± 45.6 | 749.5 ± 51.0 | 999.8 ± 98.5 | 1752.6 ± 561.4 |
| High-risk volume, n a | 5.1 ± 4.3 | 23.1 ± 12.0 | 52.9 ± 16.0 | 83.5 ± 24.0 | 119.0 ± 33.4 | 187.3 ± 53.0 | 253.6 ± 67.5 | 355.1 ± 88.4 | 466.8 ± 109.0 | 891.2 ± 332.6 |
| Low-risk volume, n a | 10.4 ± 8.8 | 41.0 ± 12.3 | 82.6 ± 19.4 | 124.6 ± 24.0 | 172.1 ± 31.6 | 234.7 ± 56.5 | 321.6 ± 61.4 | 394.4 ± 85.9 | 533.0 ± 119.2 | 861.5 ± 308.8 |
| Age, y a | 26.7 ± 5.7 | 27.2 ± 6.0 | 27.4 ± 6.2 | 27.9 ± 6.2 | 28.5 ± 6.3 | 29.0 ± 6.3 | 29.3 ± 6.4 | 29.1 ± 6.3 | 29.3 ± 6.3 | 29.9 ± 6.3 |
| Race, n (%) | ||||||||||
| White | 653 (70.8) | 3842 (64.0%) | 7552 (53.3) | 11,888 (57.2) | 13,687 (45.3) | 21,857 (48.1) | 24,980 (41.2) | 32,986 (41.8) | 40,745 (39.8) | 67,392 (38.1) |
| Hispanic | 59 (6.4) | 418 (7.0) | 1938 (13.7) | 2635 (12.7) | 6940 (23.0) | 10,554 (23.2) | 16,155 (26.6) | 21,102 (26.7) | 29,559 (28.8) | 49,018 (27.7) |
| Black | 43 (4.7) | 227 (3.8) | 775 (5.5) | 1093 (5.3) | 2748 (9.1) | 5606 (12.3) | 8000 (13.2) | 11,851 (15.0) | 9687 (9.5) | 22,147 (12.5) |
| Other | 167 (18.1) | 1519 (25.3) | 3910 (27.6) | 5163 (24.9) | 6854 (22.7) | 7418 (16.3) | 11,524 (19.0) | 13,068 (16.5) | 22,490 (22.0) | 38,523 (21.8) |
| Payor, n (%) | ||||||||||
| Medicaid | 366 (39.7) | 2893 (48.2) | 6850 (48.3) | 9119 (43.9) | 12,589 (41.7) | 16,771 (36.9) | 25,559 (42.1) | 31,104 (39.4) | 40,108 (39.1) | 65,216 (36.8) |
| Private | 449 (48.7) | 2733 (45.5) | 6237 (44.0) | 10,535 (50.7) | 15,225 (50.4) | 25,838 (56.9) | 31,530 (52.0) | 42,733 (54.1) | 56,965 (55.6) | 104,763 (59.2) |
| Medicare | 7 (0.8) | 37 (0.6) | 138 (0.1) | 112 (0.5) | 204 (0.7) | 208 (0.5) | 269 (0.4) | 378 (0.5) | 455 (0.4) | 552 (0.3) |
| Self-pay | 82 (8.9) | 225 (3.8) | 444 (3.1) | 524 (2.5) | 1443 (4.8) | 1559 (3.4) | 2135 (3.5) | 3198 (4.1) | 2855 (2.8) | 3831 (2.2) |
| Comorbidities, n (%) | ||||||||||
| Advanced maternal age | 93 (10.1) | 798 (13.3) | 2124 (15.0) | 3546 (17.1) | 5832 (19.3) | 9862 (21.7) | 13,878 (22.9) | 17,408 (22.0) | 23,115 (22.6) | 45,659 (25.8) |
| Hypertension disorder | 49 (5.3) | 380 (6.3) | 995 (7.0) | 1386 (6.7) | 2123 (7.0) | 3019 (6.6) | 4288 (7.1) | 5922 (7.5) | 7731 (7.5) | 15,330 (8.7) |
| Diabetes mellitus | 55 (6.0) | 327 (5.4) | 886 (6.3) | 1424 (6.9) | 2240 (7.4) | 3165 (7.0) | 4564 (7.5) | 6204 (7.9) | 8633 (8.4) | 15,335 (8.7) |
| Obesity | 17 (1.8) | 165 (2.8) | 331 (2.3) | 548 (2.6) | 739 (2.4) | 1137 (2.5) | 1581 (2.6) | 2275 (2.9) | 3655 (3.6) | 4810 (2.7) |
| Multiple gestations | 5 (0.5) | 34 (0.6) | 112 (0.8) | 179 (0.9) | 273 (0.9) | 515 (1.1) | 774 (1.3) | 1158 (1.5) | 1915 (1.9) | 3720 (2.1) |
| Preterm gestation | 24 (2.6) | 224 (3.7) | 537 (3.8) | 928 (4.5) | 1622 (5.4) | 2570 (5.7) | 3851 (6.4) | 5528 (7.0) | 8389 (8.2) | 16,279 (9.2) |
In analyses of unadjusted outcomes ( Table 3 ; Figure 2 ), we found higher rates of the composite adverse outcome and most of the individual adverse outcomes in the lowest volume hospitals (deciles 1 and 2), when compared with all other hospitals within all deliveries, vaginal deliveries, or cesarean deliveries. For example, with at all deliveries in aggregate, the incidence of the composite outcome in decile 1 was 11.8% and in decile 2 was 10.1%; the incidence ranged from 7.6-8.8% for the other 8 deciles in aggregate ( P < .0001). Looking at spontaneous vaginal deliveries and cesarean deliveries in isolation, we saw similar results with higher unadjusted rates of both the composite and most individual adverse outcomes in decile 1 and 2 hospitals. We also observed a modest increase in unadjusted complication rates in the highest volume hospitals (deciles 8-10) when compared with intermediate volume hospitals (deciles 5-7). This effect was particularly apparent in cesarean deliveries ( Table 3 ). To ensure the robustness of our findings, we repeated our analyses examining the relationship between volume and outcome while stratifying hospitals on the basis of total delivery volume rather than vaginal or cesarean delivery volume; thus, in these analyses, we explored the relationship between vaginal delivery outcomes and volume after stratifying hospitals by their total delivery volume and did analogous analyses for cesarean deliveries. We found that the volume-outcome relationship was similar, irrespective of the measure of volume that was used. We also calculated the proportion of cesarean deliveries according to hospital decile to cesarean delivery volume ( Table 3 ) and found no difference in the proportion of deliveries that were cesarean sections across volume deciles.
| Variable | Decile 1 | Decile 2 | Decile 3 | Decile 4 | Decile 5 | Decile 6 | Decile 7 | Decile 8 | Decile 9 | Decile 10 | P value |
|---|---|---|---|---|---|---|---|---|---|---|---|
| All deliveries | |||||||||||
| N | 3011 | 20,678 | 45,446 | 69,869 | 99,654 | 141,844 | 188,227 | 243,475 | 331,801 | 534,804 | |
| Hemorrhage, n (%) | 241 (8.0) | 1383 (6.7) | 2527 (5.6) | 3644 (5.2) | 4849 (4.9) | 6378 (4.5) | 8878 (4.7) | 12,342 (5.1) | 17,539 (5.%) | 26,210 (4.9) | < .0001 |
| Infection, n (%) | 26 (0.9) | 102 (0.5) | 212 (0.5) | 252 (0.4) | 455 (0.5) | 693 (0.5) | 1178 (0.6) | 2061 (0.9) | 2450 (0.7) | 4709 (0.9) | < .0001 |
| Laceration, n (%) | 84 (2.8) | 528 (2.6) | 927 (2.0) | 1508 (2.2) | 2152 (2.2) | 2859 (2.0) | 3404 (1.8) | 5005 (2.1) | 7161 (2.2) | 10,910 (2.0%) | < .0001 |
| Thrombotic, n (%) | 26 (0.9) | 252 (1.2) | 542 (1.2) | 874 (1.3) | 1168 (1.2) | 1674 (1.2) | 2270 (1.2) | 2,23 (1.2) | 4366 (1.3) | 7116 (1.3) | < .0001 |
| Operative, n (%) | 9 (0.3) | 58 (0.3) | 121 (0.3) | 227 (0.3) | 339 (0.3) | 426 (0.3) | 576 (0.3) | 921 (0.4) | 1,350 (0.4) | 2413 (0.5) | < .0001 |
| Mortality, n (%) | 1 (0.03) | 2 (0.01) | 2 (0.00) | 5 (0.01) | 4 (0.00) | 13 (0.01) | 17 (0.01) | 20 (0.01) | 27 (0.01) | 56 (0.01) | .5449 |
| Composite, n (%) | 354 (11.8) | 2082 (10.1) | 3880 (8.5) | 5884 (8.4) | 8021 (8.1) | 10,788 (7.6) | 14,531 (7.7) | 20,651 (8.5) | 29,151 (8.8) | 45,288 (8.5) | < .0001 |
| Vaginal deliveries | |||||||||||
| N | 1898 | 13,301 | 28,511 | 45,112 | 63,108 | 88,309 | 117,137 | 150,948 | 210,727 | 328,797 | |
| Hemorrhage, n (%) | 136 (7.2) | 806 (6.1) | 1,458 (5.1) | 2,232 (5.0) | 2,740 (4.3) | 3,732 (4.2) | 4,972 (4.2) | 6,571 (4.4) | 9,254 (4.4) | 13,540 (4.1) | < .0001 |
| Infection, n (%) | 13 (0.7) | 32 (0.2) | 89 (0.3) | 121 (0.3) | 181 (0.3) | 286 (0.3) | 472 (0.4) | 697 (0.5) | 950 (0.5) | 1707 (0.5) | < .0001 |
| Laceration, n (%) | 50 (2.6) | 354 (2.7) | 596 (2.1) | 1025 (2.3) | 1379 (2.2) | 1922 (2.2) | 2202 (1.9) | 3378 (2.2) | 4707 (2.2) | 7016 (2.1) | < .0001 |
| Thrombotic, n (%) | 6 (0.3) | 41 (0.3) | 93 (0.3) | 168 (0.4) | 228 (0.4) | 312 (0.4) | 378 (0.3) | 470 (0.3) | 692 (0.3) | 1063 (0.3) | .5073 |
| Mortality, n (%) | 0 (0.00) | 0 (0.00) | 0 (0.00) | 0 (0.00) | 0 (0.00) | 4 (0.00) | 6 (0.01) | 7 (0.00) | 3 (0.00) | 21 (0.01) | .0868 |
| Composite, n (%) | 194 (10.2) | 1,150 (8.7) | 2,123 (7.5) | 3,379 (7.5) | 4,283 (6.8) | 5,935 (6.7) | 7,617 (6.5) | 10,551 (7.0) | 14,771 (7.0) | 22,126 (6.7) | < .0001 |
| Cesarean deliveries | |||||||||||
| N | 922 | 6006 | 14,175 | 20,779 | 30,229 | 45,435 | 60,659 | 79,007 | 102,481 | 177,080 | |
| Hemorrhage, n (%) | 86 (9.3) | 438 (7.3) | 841 (5.9) | 1103 (5.3) | 1665 (5.5) | 2115 (4.7) | 3264 (5.4) | 4958 (6.3) | 7140 (7.0) | 10,819 (6.1) | < .0001 |
| Infection, n (%) | 12 (1.3) | 55 (0.9) | 102 (0.7) | 111 (0.5) | 246 (0.8) | 351 (0.8) | 633 (1.0) | 1231 (1.6) | 1360 (1.3) | 2779 (1.6) | < .0001 |
| Operative, n (%) | 9 (1.0) | 58 (1.0) | 121 (0.9) | 227 (1.1) | 339 (1.1) | 426 (0.9) | 576 (1.0) | 921 (1.2) | 1350 (1.3) | 2413 (1.4%) | < .0001 |
| Thrombotic, n (%) | 11 (1.2) | 70 (1.2) | 182 (1.3) | 249 (1.2) | 363 (1.2) | 493 (1.1) | 803 (1.3) | 1152 (1.5) | 1578 (1.5) | 2881 (1.6) | < .0001 |
| Mortality, (n) | 1 (0.11) | 2 (0.03) | 2 (0.01) | 3 (0.01) | 4 (0.01) | 9 (0.02) | 11 (0.02) | 13 (0.02) | 22 (0.02) | 34 (0.02) | .7240 |
| Composite, n (%) | 103 (11.2) | 535 (8.9) | 1043 (7.4) | 1412 (6.8) | 2166 (7.2) | 2809 (6.2) | 4342 (7.2) | 6810 (8.6) | 9378 (9.2) | 15,304 (8.6) | < .0001 |
In analyses that adjusted for patient demographics and comorbidities ( Table 4 ), we again found higher risk of adverse outcomes in lowest volume hospitals (decile 1 and 2). Specifically, odds of experiencing the composite outcome were between 43% and 60% higher for decile 1 hospitals (with decile 10 serving as the reference category) even after adjustment for patient demographics and comorbidity ( P < .05). Alternatively, the odds of adverse outcomes appeared modestly lower for intermediate volume hospitals (decile 5 and 6), with the effect particularly notable for cesarean deliveries.
| Lowest volume | Highest volume | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Decile 1 | Decile 2 | Decile 3 | Decile 4 | Decile 5 | Decile 6 | Decile 7 | Decile 8 | Decile 9 | Decile 10 | ||||||||||
| Adverse outcomes | OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | |
| All deliveries | Reference | ||||||||||||||||||
| Unadjusted | 1.37 (1.07–1.75) | .0112 | 1.17 (0.98–1.39) | .0901 | 0.92 (0.78–1.10) | .3712 | 0.96 (0.81–1.14) | .6641 | 0.87 (0.73–1.03) | .1046 | 0.81 (0.68–0.96) | .0142 | 0.86 (0.73–1.02) | .0814 | 0.96 (0.81–1.14) | .6643 | 0.97 (0.82–1.15) | .7279 | |
| Adjusted for demographics a | 1.38 (1.08–1.76) | .0100 | 1.17 (0.98–1.40) | .0861 | 0.92 (0.77–1.10) | .3333 | 0.96 (0.81–1.14) | .6481 | 0.87 (0.73–1.02) | .0956 | 0.81 (0.69–0.96) | .0155 | 0.86 (0.73–1.02) | .0819 | 0.96 (0.81–1.14) | .6622 | 0.97 (0.82–1.14) | .6867 | |
| Adjusted for demographics and selected comorbidities b | 1.43 (1.13–1.82) | .0035 | 121 (1.02–1.44) | .0328 | 0.95 (0.80–1.12) | .5304 | 0.99 (0.83–1.17) | .8652 | 0.89 (0.75–1.05) | .1640 | 0.83 (0.70–0.98) | .0277 | 0.87 (0.74–1.03) | .1112 | 0.98 (0.83–1.15) | .7674 | 0.97 (0.82–1.15) | .7322 | |
| Vaginal deliveries | Reference | ||||||||||||||||||
| Unadjusted | 1.51 (1.16–1.96) | .0024 | 1.28 (1.08–1.53) | .0055 | 1.05 (0.88–1.24) | .6006 | 1.10 (0.93–1.30) | .2825 | 0.94 (0.79–1.11) | .4364 | 0.90 (0.76–1.06) | .1940 | 0.92 (0.78–1.08) | .2907 | 1.00 (0.85–1.17) | .9625 | 0.98 (0.83–1.16) | .8179 | |
| Adjusted for demographics a | 1.57 (1.13–1.92) | < .0001 | 1.32 (1.10–1.57) | < .0001 | 1.05 (0.89–1.24) | .0002 | 1.11 (0.94–1.32) | < .0001 | 0.92 (0.78–1.09) | .9741 | 0.91 (0.77–1.08) | .7261 | 0.92 (0.78–1.08) | .0066 | 0.99 (0.84–1.17) | .0014 | 0.99 (0.84–1.16) | .0004 | |
| Adjusted for demographics and selected comorbidities | 1.60 (1.38–1.86) | < .0001 | 1.33 (1.25–1.42) | < .0001 | 1.12 (1.07–1.17) | < .0001 | 1.13 (1.08–1.17) | < .0001 | 1.02 (0.99–1.06) | .2401 | 1.01 (0.98–1.04) | .3933 | 0.97 (0.95–1.00) | .0581 | 1.05 (1.02–1.07) | .0002 | 1.04 (1.02–1.06) | .0002 | |
| Cesarean deliveries | Reference | ||||||||||||||||||
| Unadjusted | 1.37 (0.96–1.95) | .0828 | 1.01 (0.79–1.29) | .9507 | 0.79 (0.62–1.00) | .0469 | 0.76 (0.60–0.95) | .0175 | 0.73 (0.58–0.91) | .0063 | 0.65 (0.52–0.82) | .0002 | 0.77 (0.62–0.97) | .0252 | 0.94 (0.75–1.18) | .5991 | 0.94 (0.75–1.18) | .5986 | |
| Adjusted for demographics a | 1.38 (0.97–1.97) | .0760 | 1.02 (0.80–1.31) | .8841 | 0.79 (0.62–1.00) | .0462 | 0.76 (0.61–0.96) | .0224 | 0.73 (0.58–0.91) | .0062 | 0.65 (0.52–0.82) | .0002 | 0.77 (0.61–0.96) | .0216 | 0.93 (0.75–1.17) | .5482 | 0.94 (0.75–1.18) | .6020 | |
| Adjusted for demographics and selected comorbidities b | 1.45 (1.01–2.04) | .0436 | 1.06 (0.83–1.36) | .6286 | 0.82 (0.65–1.03) | .0871 | 0.79 (0.63–0.99) | .0402 | 0.75 (0.60–0.94) | .0129 | 0.67 (0.53–0.84) | .0004 | 0.78 (0.62–0.97) | .0278 | 0.95 (0.76–1.18) | .6250 | 0.95 (0.76–1.18) | .6313 | |
a Adjusted for race, age, and payor;
b Adjusted for race, age, payor, advanced age, herpes, asthma, cerebral hemorrhage, chorioamnionitis, diabetes mellitus, hypertensive disorders, congenital heart disease, liver anomalies, renal anomalies, thyroid disease, mental disorder, multiple gestation, preterm gestation, obesity, pulmonary embolism, and uterine rupture.
To ensure the robustness of our findings, we conducted additional supplemental analyses. First, we repeated our unadjusted and adjusted analyses examining outcomes across deciles of hospital volume after stratifying patients into high- and low-risk subgroups. Unadjusted analyses produced similar results to our main analyses, with higher complication rates in the lowest volume hospitals for both the high- and low-risk patient strata, irrespective of delivery route ( Appendix; Supplementary Tables 1-3 ). Adjusted results in the high- and low-risk subgroups were also similar to the main analyses, with higher odds of experiencing the composite outcomes in the lowest volume hospitals ( Appendix; Supplementary Tables 4-6 ). Additional results were significant for high-risk patients who had higher odds of adverse outcomes compared with low-risk subgroups in the lowest volume hospitals ( P < .05). Interestingly, in our supplementary analyses high-volume hospitals again appeared to have marginally higher rates of adverse outcomes in both the high- and low-risk strata. We also conducted additional analyses using alternative methods for the stratification of hospitals according to volume (eg, quartiles, quintiles), and the results were again similar.
Results
We identified a total of 1,683,754 childbirths. After excluding transfer cases (n = 4945), our final cohort of spontaneous vaginal deliveries, forceps deliveries, vacuum extractions, and cesarean deliveries included 1,678,809 admissions to 1045 hospitals. After the exclusion of forceps and vacuum-assisted deliveries (n = 94,188), there were a total of 1,047,848 spontaneous vaginal deliveries in 1011 hospitals (34 hospitals delivered only by cesarean section) and 536,773 cesarean deliveries in 1030 hospitals (15 hospitals delivered only by vaginal route). For hospitals that performed both vaginal and cesarean deliveries, the cesarean delivery rate ranged from 13.0–96.7% across 1030 hospitals (mean, 31.3%; median, 30.3%). The mean hospital childbirth volume was 1606.5 (interquartile range, 442.0–2299.0); the mean vaginal delivery volume was 1036.4 (interquartile range, 308.0–1472.0), and the mean cesarean delivery volume was 521.1 (interquartile range, 135–754).
The characteristics of patients with spontaneous vaginal delivery that were stratified by hospital volume deciles are displayed in Table 1 . Women who delivered in lower volume hospitals tended to younger, were more likely to be white, and were more likely to be categorized as self-pay when compared with women who delivered at higher volume hospitals. Women who underwent vaginal delivery in lower volume hospitals tended to have fewer comorbid conditions such as advanced maternal age, hypertension, and diabetes mellitus ( Table 1 ). Conversely, women who had vaginal deliveries in high-volume hospitals were more likely to be older and Hispanic or black and to have more comorbid illness. Results were similar for cesarean deliveries ( Table 2 ), with low-volume hospitals treating a higher proportion of uninsured younger white women, although high-volume hospitals treated more women with advanced age and comorbid illness.
