Risk-adjusted models for adverse obstetric outcomes and variation in risk-adjusted outcomes across hospitals

Research

www.AJOG .org

OBSTETRICS

Risk-adjusted models for adverse obstetric outcomes and variation in risk-adjusted outcomes across hospitals Jennifer L. Bailit, MD, MPH; William A. Grobman, MD, MBA; Madeline Murguia Rice, PhD; Catherine Y. Spong, MD; Ronald J. Wapner, MD; Michael W. Varner, MD; John M. Thorp, MD; Kenneth J. Leveno, MD; Steve N. Caritis, MD; Phillip J. Shubert, MD; Alan T. Tita, MD, PhD; George Saade, MD; Yoram Sorokin, MD; Dwight J. Rouse, MD; Sean C. Blackwell, MD; Jorge E. Tolosa, MD, MSCE; J. Peter Van Dorsten, MD; for the Eunice Kennedy Shriver National Institute of Child Health and Human Development MaternaleFetal Medicine Units Network OBJECTIVE: Regulatory bodies and insurers evaluate hospital quality using obstetrical outcomes, however meaningful comparisons should take preexisting patient characteristics into account. Furthermore, if risk-adjusted outcomes are consistent within a hospital, fewer measures and resources would be needed to assess obstetrical quality. Our objective was to establish risk-adjusted models for 5 obstetric outcomes and assess hospital performance across these outcomes. STUDY DESIGN: We studied a cohort of 115,502 women and their

neonates born in 25 hospitals in the United States from March 2008 through February 2011. Hospitals were ranked according to their unadjusted and risk-adjusted frequency of venous thromboembolism, postpartum hemorrhage, peripartum infection, severe perineal laceration, and a composite neonatal adverse outcome. Correlations between hospital risk-adjusted outcome frequencies were assessed. RESULTS: Venous thromboembolism occurred too infrequently

(0.03%; 95% confidence interval [CI], 0.02e0.04%) for meaningful

assessment. Other outcomes occurred frequently enough for assessment (postpartum hemorrhage, 2.29%; 95% CI, 2.20e2.38, peripartum infection, 5.06%; 95% CI, 4.93e5.19, severe perineal laceration at spontaneous vaginal delivery, 2.16%; 95% CI, 2.06e2.27, neonatal composite, 2.73%; 95% CI, 2.63e2.84). Although there was high concordance between unadjusted and adjusted hospital rankings, several individual hospitals had an adjusted rank that was substantially different (as much as 12 rank tiers) than their unadjusted rank. None of the correlations between hospital-adjusted outcome frequencies was significant. For example, the hospital with the lowest adjusted frequency of peripartum infection had the highest adjusted frequency of severe perineal laceration. CONCLUSION: Evaluations based on a single risk-adjusted outcome cannot be generalized to overall hospital obstetric performance.

Key words: obstetrics, performance improvement, quality, risk adjustment

Cite this article as: Bailit JL, Grobman WA, Rice MM, et al. Risk-adjusted models for adverse obstetric outcomes and variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013;209:446.e1-30.

From the Departments of Obstetrics and Gynecology, Case Western Reserve UniversityeMetroHealth Medical Center, Cleveland, OH (Dr Bailit); Northwestern University, Chicago, IL (Dr Grobman); Columbia University, New York, NY (Dr Wapner); University of Utah Health Sciences Center, Salt Lake City, UT (Dr Varner); University of North Carolina at Chapel Hill, Chapel Hill, NC (Dr Thorp); University of Texas Southwestern Medical Center, Dallas, TX (Dr Leveno); University of Pittsburgh, Pittsburgh, PA (Dr Caritis); St Ann’s Hospital, Ohio State University, Columbus, OH (Dr Shubert); University of Alabama at Birmingham, Birmingham, AL (Dr Tita); University of Texas Medical Branch, Galveston, TX (Dr Saade); Wayne State University, Detroit, MI (Dr Sorokin); Brown University, Providence, RI (Dr Rouse); University of Texas Health Science Center at Houston-Children’s Memorial Hermann Hospital, Houston, TX (Dr Blackwell); Oregon Health and Science University, Portland, OR (Dr Tolosa); Medical University of South Carolina, Charleston, SC (Dr Van Dorsten); and George Washington University Biostatistics Center, Washington, DC (Dr Rice); and the Eunice Kennedy Shriver National Institute of Child Health and Human Development, Bethesda, MD (Dr Spong). The remaining members of the MaternaleFetal Medicine Units Network appear in the Acknowledgments. Received April 19, 2013; revised June 30, 2013; accepted July 22, 2013. The project described was supported by grants from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) (HD21410, HD27869, HD27915, HD27917, HD34116, HD34208, HD36801, HD40500, HD40512, HD40544, HD40545, HD40560, HD40485, HD53097, HD53118) and the National Center for Research Resources (NCRR) (UL1 RR024989; 5UL1 RR025764) and its contents do not necessarily represent the official views of the NICHD, NCRR, or National Institutes of Health. The authors report no conflict of interest. Presented at the 32nd annual meeting of the Society for Maternal-Fetal Medicine, Dallas, TX, Feb. 6-11, 2012. Reprints are not available from the authors. 0002-9378/$36.00
ª 2013 Mosby, Inc. All rights reserved.
http://dx.doi.org/10.1016/j.ajog.2013.07.019

446.e1 American Journal of Obstetrics & Gynecology NOVEMBER 2013

Obstetrics

www.AJOG.org

A

dmission for delivery constitutes the most common indication for hospitalization in the United States.1 The sheer volume of deliveries, as well as the fact that each admission has the potential to affect the short- and long-term health of at least 2 individuals (a mother and her newborn), underscores the importance of achieving high-quality delivery care. Correspondingly, measuring these outcomes should be an important component of quality improvement.1-3 An ideal quality measure for inpatient obstetrics would encompass 5 major characteristics: (1) association with meaningful maternal and neonatal outcomes, (2) relation to outcomes that are influenced by physician/health system behaviors, (3) affordability for application on a large-scale basis, (4) acceptability to practicing obstetricians as a meaningful marker of quality, and (5) reliability/ reproducibility.4 Yet, because outcomes may be dependent upon preexisting patient characteristics, simply measuring health outcomes may not provide insight into quality of care or allow valid comparisons among institutions. To overcome this limitation, risk adjustment has been widely employed in clinical disciplines such as cardiothoracic surgery to assess outcomes for procedures such as lung resection or coronary artery bypass grafting.5-7 Such risk adjustment, however, has been used inconsistently in evaluation of obstetric outcomes8 and there is no consensus on which outcomes or which risk adjustment factors should be used. Moreover, as quality measurement increases, so does the need to do such measurement parsimoniously. Many consumers and providers tend to think of hospitals as having consistent quality within a given discipline, in which case measures should be highly correlated and fewer aspects of care would need to be assessed. Conversely, if measures are not correlated, multiple measures would need to be collected to enable an accurate assessment of performance. Our objective was thus to establish risk adjustment models for use in obstetrics that adjust for preexisting patient characteristics and, using these models, assess the consistency of hospital performance across obstetrical outcomes

Research

TABLE 1

Maternal (n [ 115,502) and neonatal (n [ 118,422) characteristics of study population Characteristics

No. (%)

MATERNAL CHARACTERISTICS Age, y <20

10,187 (8.8)

20-24.9

24,299 (21.0)

25-29.9

31,101 (26.9)

30-34.9

30,570 (26.5)

35

19,345 (16.8) a

Race/ethnicity

Non-Hispanic white

52,040 (45.1)

Non-Hispanic black

23,878 (20.7)

Non-Hispanic Asian

5999 (5.2)

Hispanic

27,291 (23.6)

Other

5083 (4.4)

Not documented

1211 (1.1) b

Body mass index at delivery, kg/m

2

<25

14,242 (12.6)

25-29.9

41,268 (36.5)

30-34.9

32,088 (28.4)

35-39.9

15,088 (13.3)

40

10,481 (9.3)

Cigarette use during pregnancy Cocaine or methamphetamine use during pregnancy

11,370 (9.9) 830 (0.7)

Insurance status Uninsured/self-pay

11,989 (10.5)

Government-assisted

45,125 (39.4)

Private

57,462 (50.2)

Prenatal careb

107,510 (97.9)

Obstetric history Nulliparous

46,773 (40.5)

Prior vaginal delivery only

49,865 (43.2)

Prior cesarean only

8872 (7.7)

Prior cesarean and vaginal

9963 (8.6)

Any hypertension

13,272 (11.5)

Diabetes mellitus None

106,706 (92.4)

Gestational

6999 (6.1)

Pregestational

1734 (1.5)

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

(continued)

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e2

Research

Obstetrics

that fit the criteria listed above for quality outcome measures.

M ATERIALS

AND

M ETHODS

Study design From 2008-2011, we assembled a cohort of women and their neonates born in any of the 25 hospitals of the Eunice Kennedy Shriver National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network. The Assessment of Perinatal Excellence study was designed to develop quality measures for intrapartum obstetrical care. This study was approved by the institutional review board at each participating institution under a waiver of informed consent. Patients eligible for data collection were those who delivered within the institution, were at least 23 weeks of gestation, and had a live fetus on admission. Data were collected on eligible patients if they delivered during the 24-hour period of selected days during a 3-year period (March 2008 February 2011). Days were chosen via computer-generated random selection. To avoid overrepresentation of patients from larger hospitals, we selected one third of days at hospitals with annual delivery volumes from 2000-7000 and up to one sixth of days at hospitals with annual deliveries >7000. The randomization scheme was stratified by weekdays, weekends, and holidays and generated separately for each hospital. On selected days, the labor and delivery logbook at each participating center was screened to identify all eligible women. The medical records of all eligible women and their newborns were abstracted by trained and certified research personnel at the hospital and entered into a World Wide Webebased data entry system. Data recorded included demographic characteristics, details of the medical and obstetrical history, information about intrapartum and postpartum events, and patients’ race and ethnicity as reported in the chart. Maternal data were collected until discharge and neonatal data were collected up until discharge or until 120 days of age, whichever came first. Feasibility and quality of data collection were ensured by several mechanisms.

www.AJOG.org

TABLE 1

Maternal (n [ 115,502) and neonatal (n [ 118,422) characteristics of study population (continued) Characteristics Anticoagulant use during pregnancy Multiple gestation

No. (%) 920 (0.8) 2815 (2.4)

Polyhydramnios

940 (0.8)

Oligohydramnios

4700 (4.1)

Placenta previa

467 (0.4)

Placenta accreta

158 (0.1)

Placental abruption

930 (0.8)

b

PROM/PPROM

6004 (5.3)

GBS status Negative

68,918 (59.7)

Positive

24,390 (21.1)

Unknown

22,194 (19.2)

NEONATAL CHARACTERISTICS Presentation at delivery Vertex

111,174 (94.1)

Breech

6010 (5.1)

Nonbreech malpresentation

931 (0.8)

Gestational age at delivery, wk 230-276

1256 (1.1)

0

6

4282 (3.6)

0

6

34 -36

10,024 (8.5)

370-376

10,914 (9.2)

28 -33

0

6

38 -38

20,723 (17.5)

0

6

39 -39

37,695 (31.8)

0

6

40 -40

23,876 (20.2)

0

6

8998 (7.6)

41 -41 42

0

654 (0.6)

Birthweight, g <2500

12,498 (10.6)

2500-3999

96,708 (81.7)

4000

9186 (7.8)

Size for gestational age Small

11,530 (9.7)

Appropriate

97,774 (82.6)

Large

9088 (7.7)

GBS, group B streptococcus; PPROM, preterm premature rupture of membranes; PROM, premature rupture of membranes. a

Race/ethnicity was reported in chart; b n ¼ 113,167 with body mass index data, n ¼ 109,773 with prenatal care visit data, n ¼ 113,446 with PROM/PPROM data.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

446.e3 American Journal of Obstetrics & Gynecology NOVEMBER 2013

Obstetrics

www.AJOG.org

Research

TABLE 2

Observed hospital frequencies of obstetric outcomes No. of outcomes

Outcome Venous thromboembolisma Postpartum hemorrhage Peripartum infection

b

b

Denominator size for each outcome

Lowest frequency percent

Frequency percent (95% CI)

Median frequency percent

Highest frequency percent

31

115,499

0.03 (0.02e0.04)

0.00

0.02

0.07

2425

105,987

2.29 (2.20e2.38)

0.82

2.09

4.86

5581

110,205

5.06 (4.93e5.19)

2.19

5.34

9.69

Severe perineal laceration at SVDc

1475

68,144

2.16 (2.06e2.27)

1.01

2.00

4.89

c

523

1898

27.56 (25.54e29.57)

8.00

32.56

48.15

510

3515

14.51 (13.34e15.67)

3.73

13.99

48.15

2440

89,279

2.73 (2.63e2.84)

0.96

2.61

5.91

Severe perineal laceration at FVD

Severe perineal laceration at VVD

c

Composite neonatal adverse outcome

d

CI, confidence interval; FVD, forceps-assisted vaginal delivery; SVD, spontaneous vaginal delivery; VVD, vacuum-assisted vaginal delivery. a

Among all women with complete outcome data; b Among all women with complete outcome and covariable data; c Among women with singleton delivery and no shoulder dystocia or placenta previa and complete outcome and covariable data; d Among women with term, nonanomalous singleton infants and complete outcome and covariable data.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

First, prior to selecting final data fields and forms, a 2-week pilot study took place to evaluate the data collection process, quality of the data, and frequency of missing data. Based on the information gathered during this pilot phase, final data fields were selected and forms revised. All data were subjected to ongoing data edits to ensure accuracy.

Primary outcomes An initial determination of the primary obstetric outcomes of interest was made via expert consensus, obtained during meetings of members of the MaternalFetal Medicine Units Steering Committee and an external advisory committee convened specifically for this project (Acknowledgments). Based on input from these committees, 5 primary outcomes were chosen because they represented different domains of obstetric complications, were clinically meaningful, could be affected by differences in clinical care, were ascertainable from medical records, and potentially occurred with sufficient frequency to allow valid institutional comparisons: venous thromboembolism, postpartum hemorrhage, peripartum infection, severe perineal laceration, and a composite neonatal adverse outcome.9-15 Venous thromboembolism was defined as occurrence of either a deep venous thrombosis diagnosed by duplex Doppler or a pulmonary embolism diagnosed by

computed tomography or ventilationperfusion lung scan. Postpartum hemorrhage was defined as occurrence of any of the following: an estimated blood loss 1500 mL at delivery or the immediate postpartum period, a blood transfusion, or a hysterectomy for hemorrhage, placenta accreta, or atony. Peripartum infection was defined as occurrence of any of the following: chorioamnionitis, endometritis, wound cellulitis requiring antibiotics, wound reopened for fluid collection or infection, or wound dehiscence during the delivery hospitalization. Severe perineal laceration was defined as the occurrence of a third- or fourthdegree perineal laceration, was restricted to women with vaginal singleton deliveries with no shoulder dystocia or placenta previa, and was stratified by spontaneous, vacuum, or forceps delivery. The composite neonatal adverse outcome was defined as occurrence of any of the following restricted to term (37 weeks of gestation), nonanomalous singleton infants: neonatal stay longer than maternal stay by 3 calendar days, 5-minute Apgar score <4, skeletal fracture other than of the clavicle, facial nerve palsy, brachial plexus palsy, subgaleal hemorrhage, ventilator support, hypoxic ischemic encephalopathy, stillbirth after hospital admission, or neonatal death. We chose to include only term infants because morbidity prior to term is strongly related to prematurity-related complications and

not necessarily differences in health care and thus inclusion is not as good a candidate for an obstetric quality measure. Additional details regarding the definitions of these outcomes and relevant denominators can be found in the Appendix; Supplementary Table 1.

Statistical analyses At each institution, the unadjusted frequencies of adverse outcomes, with 95% confidence intervals (CIs), were calculated and were compared using the c2 test. The analysis was then directed at assessing which patient characteristics were significantly associated with the chosen outcomes. We chose not to examine hospital characteristics such as volume or teaching status in the model as they are not inherently related to the patient and should not be included in a risk-adjustment model. Patient characteristics eligible for multivariable models were selected a priori based on whether they could plausibly be associated with the outcome. While inclusion of the same risk factors in all models would be convenient, requiring this condition could lead to loss of precision of the risk-adjustment models that were developed. As our goal was to have the best possible risk-adjustment models, we included as potential factors for each model the most relevant and plausible risk factors. Prior to multivariable analysis, the possibility of colinearity

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e4

Final multivariable model for each obstetric outcome Variable

Postpartum hemorrhagea

Peripartum infectiona

Severe perineal laceration at SVDb

Severe perineal laceration at FVDb,c

Severe perineal laceration at VVDb,c

Composite neonatal adverse outcomed

Denominator size

105,987

110,205

68,144

1898

3515

89,279

C

C

C

C

C

Body mass index at delivery

C

C

C

C

C

Cigarette use during pregnancy

C

C

C

C

C

Maternal characteristics

Obstetrics

Age

Research

446.e5 American Journal of Obstetrics & Gynecology NOVEMBER 2013

TABLE 3

C

Cocaine or methamphetamine use during pregnancy Insurance status

C

Prenatal care

C

Obstetric history

C

Any hypertension

C

Diabetes mellitus (gestational, pregestational)

C

C

C

C

C

C C

C

C

C

C

C C

C

C

PROM/PPROM

C

C

GBS status

C

Anticoagulant use during pregnancy C Multiple gestation

C

Placenta previa

C

Placenta accreta

C

Placental abruption

C

Neonatal characteristics C

Gestational age at delivery Birthweight

C

C

C

C C

Size for gestational age C statistic (95% CI)

0.74 (0.73e0.75)

0.75 (0.74e0.75)

0.79 (0.78e0.80)

0.68 (0.65e0.70)

0.69 (0.67e0.72)

0.68 (0.67e0.69)

CI, confidence interval; FVD, forceps-assisted vaginal delivery; GBS, group B streptococcus; PPROM, preterm premature rupture of membranes; PROM, premature rupture of membranes; SVD, spontaneous vaginal delivery; VVD, vacuum-assisted vaginal delivery. a

Among all women with complete outcome and covariable data; b Among women with singleton delivery and no shoulder dystocia or placenta previa and complete outcome and covariable data; c Final model based on k-fold analysis for outcome of severe perineal laceration at SVD; d Among women with term, nonanomalous singleton infants and complete outcome and covariable data.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

www.AJOG.org

Dots signify that variable is in final multivariable model.

Obstetrics

www.AJOG.org

FIGURE 1

Association between hospital ranks based on observed (unadjusted) outcome frequencies and hospital ranks based on adjusted outcome frequencies

A, Postpartum hemorrhage. B, Peripartum infection. Severe perineal laceration at: C, spontaneous vaginal delivery (SVD); D, forceps-assisted vaginal delivery (FVD); and E, vacuum-assisted vaginal delivery (VVD). F, Composite neonatal adverse outcome. Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

among patient characteristics was assessed. Continuous variables were first assessed to determine whether their association with each outcome was linear, by assessing the linearity of the log (odds), using a locally weighted scatterplot smoothing technique. When there was evidence of nonlinearity, we included both linear and quadratic terms. Model selection was based on creating derivation and validation datasets using a k-fold cross-validation approach in which the cohort was randomly divided into 10 equal parts and logistic regression models, using backward selection, were generated utilizing every possible combination of 9 of the 10 sets.16 Variables with P < .05 were

retained, and each of the 10 subsamples was used for validation. The C statistic was computed to assess each model’s predictive ability (discrimination). Only those variables that were present in the logistic regression model with the highest C statistic and also were present in at least 8 of the 10 k-fold logistic regression models were chosen for the final multivariable model that included the entire dataset. Because assessment of the Hosmer-Lemeshow test statistic (P value) is not recommended for datasets as large as ours,17 model fit was assessed from graphical displays of the observed and expected number of patients within each partition of the Hosmer-Lemeshow test.

Research

The final multivariable models were then used to estimate hospitals’ expected outcome frequencies. To estimate a hospital’s expected outcome frequency, which is the hospital’s outcome frequency that would be expected given the characteristics of their patients, the predicted outcome probability was estimated for each patient and then all patient probabilities within the same hospital were averaged (see eStatistics text). These expected outcome frequencies were used to calculate an observed (unadjusted) to expected ratio (OER).18 Bootstrapping was performed on 1000 samples with replacement to estimate 99% CI around the OER and identify the hospitals that were significantly different from an OER of 1.0. OERs can be interpreted as such: if the ratio is <1.0, the hospital has fewer adverse outcomes than expected; if the ratio ¼ 1.0, the hospital has as many adverse outcomes as expected; and if the ratio is >1.0, the hospital has more adverse outcomes than expected. Because we were estimating individual hospital frequencies, the primary models did not adjust for hospital; however, regressions accounting for patient clustering within a hospital (ie, adding hospital as a fixed effect to the logistic model or as a random effect to a hierarchical model) were performed to evaluate whether either adjustment altered the strength and precision of the estimated odds ratios (ORs) for the patient characteristics. For each outcome, hospitals were ranked according to their unadjusted frequency and reranked according to their adjusted frequency, and Kendall coefficient of concordance was used to assess the degree to which these rankings were similar. Correlations of hospitaladjusted frequencies for each pair of outcomes were tested using Spearman rank correlation. SAS software (SAS Institute, Cary, NC) was used for the analyses. All tests were 2-tailed. P < .01 was used to define statistical significance and 99% CIs were estimated when directly testing a hypothesis, ie, correlations between outcomes, concordance between unadjusted and adjusted ranks, and to

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e6

Research

Obstetrics

identify hospital outliers. P < .05 and 95% CIs were estimated for model building and more descriptive analyses.

www.AJOG.org

FIGURE 2

Hospital rankings by observed (unadjusted) to expected ratio

R ESULTS During the study period, data were collected on 115,502 women and their neonates at 25 hospitals. The majority of hospitals were teaching hospitals (22/25, 88%). Most also had round-the-clock availability of a maternal-fetal medicine specialist (21/25, 84%), in-house obstetric attending (21/25, 84%), neonatologist (20/25, 80%), and dedicated obstetric anesthesiologist (22/25, 88%). The median number of deliveries at the study hospitals was 4252. Over 40% of women were nulliparous, 2.4% had a multiple gestation, and 27.4% of multiparous women had previously undergone cesarean delivery (Table 1; Supplementary Table 1 for definitions); 94.1% of newborns were vertex at delivery, 13.1% were preterm (<37 weeks’ gestation at delivery), and 10.6% weighed <2500 g at birth. Given the infrequency of venous thromboembolism it was excluded from further analysis. The frequencies of the other chosen outcomes were more common and differed significantly across hospitals (Table 2) (P < .001 for all). The variables retained in the final multivariable model for each outcome are listed in Table 3 (Supplementary Table 2 for a full list of variables assessed; Supplementary Tables 3-8 for the parameter estimates, OR, and 95% CI). A core group of patient-specific factors (maternal age, body mass index, insurance status, gestational age or birthweight, obstetric history, diabetes mellitus, and smoking) was significantly associated with multiple outcomes. The C statistic for each model, which ranged from 0.68e0.79 with lower bounds of the 95% CIs all >0.50 (Table 3), demonstrate that in all cases patient factors were at least somewhat, but not fully predictive of outcomes. Model calibration showed good model fit of the observed and expected number of patients within each partition of the Hosmer-Lemeshow test with or without each outcome (Supplementary Figures 1-6). Model fit was similar

Green indicates upper bound of OER 99% CI <1.0; red indicates lower bound of OER 99% CI >1.0;white indicates OER 99% CI includes 1.0. CI, confidence interval; OER, observed to expected ratio; SVD, spontaneous vaginal delivery. a Among all women with complete outcome and covariable data; b Among women with singleton delivery and no shoulder dystocia or placenta previa and complete outcome and covariable data; c Among women with term, nonanomalous singleton infants and complete outcome and covariable data.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

whether continuous variables were entered into the model as categorical variables based on clinically relevant cutpoints, and confirmed as appropriate from the locally weighted scatterplot smoothing plots, or as linear (and quadratic when appropriate) terms; for ease of interpretation the models with categorical variables are presented in Supplementary Tables 3-8. Overall, the OR and 95% CI associated with each patient characteristic were not substantially altered after accounting for patient clustering within a hospital in either logistic or hierarchical regression models (Supplementary Tables 3-8). The graphs of the hospital ranks based on unadjusted frequencies compared with the ranks based on adjusted frequencies are presented in Figure 1. Statistically there was a relatively high concordance between the unadjusted and adjusted ranks (Kendell coefficient of

446.e7 American Journal of Obstetrics & Gynecology NOVEMBER 2013

concordance 0.86-0.98) (Supplementary Table 9). However, there were hospitals where their rank based on their adjusted frequency differed substantially (as much as 12 rank tiers) from their rank based on their unadjusted frequency. None of the comparisons of hospital risk-adjusted frequencies between outcomes was significantly correlated: hemorrhage vs neonatal (rho ¼ e0.05, P ¼ .83), hemorrhage vs infection (rho ¼ 0.26, P ¼ .21), hemorrhage vs laceration (rho ¼ e0.29, P ¼ .16), infection vs laceration (rho ¼ e0.23, P ¼ .26), infection vs neonatal (rho ¼ 0.02, P ¼ .93), and laceration vs neonatal (rho ¼ e0.13, P ¼ .52). For each outcome, several hospitals were noted to have OERs that were significantly different from 1, a fact which indicates that they were achieving outcome frequencies that were significantly different (better or worse) than

Obstetrics

www.AJOG.org expected based on their population of patients (Figure 2). When hospitals were ranked according to their OERs and characterized by outlier status for each outcome (Figure 2, color green indicating upper bound of the OER 99% CI <1.0; color red indicating lower bound of the OER 99% CI >1.0), there was no evidence that particular hospitals consistently performed either better or worse than expected across the outcomes.

C OMMENT In this study, we developed and applied risk-adjustment models for clinically meaningful obstetric outcomes. Of the 5 outcomes that were chosen a priori, 4 (postpartum hemorrhage, peripartum infection, severe perineal laceration, and the composite neonatal adverse outcome) were found both to be frequent enough and to vary sufficiently among hospitals that they could be recommended as useful outcome measures. However, these outcomes are significantly related to multiple patient characteristics, and we believe that risk adjustment is preferable if such outcomes are used to reflect institutional quality of care. Furthermore, these risk adjustment models reveal that the rankings according to obstetric outcomes are poorly correlated with one another. Thus, performance assessment based on a single outcome measure cannot be generalized to characterize overall quality of obstetric care. Instead, multiple markers of quality need to be assessed and reported to gain insight into the obstetrical quality profile of a hospital. The Assessment of Perinatal Excellence study represents a significant contribution in the examination of obstetric outcomes among hospitals as the data are not derived from an administrative dataset but by a chart review by trained abstractors. We optimized data quality by employing a data abstraction approach, described by Pronovost et al,19 in that we used explicit definitions for data fields, standardized data collection tools, and pilot testing of data collection methods. Additionally, there were ongoing data edits. The availability of electronic medical records data should make data collection easier in the future

and represent a large improvement over trying to use the easily available but poor-quality administrative data that have been used in the past. Our analysis reveals that multiple patient factors are associated with our chosen outcomes. The C statistics are consistent with those associated with other accepted risk-adjustment models, such as those used for coronary artery bypass grafterelated mortality.5,7,20 While the overall concordance between institutions’ observed (unadjusted) and adjusted ranks was high, individual institutions may appear quite differently in terms of rank order if their adjusted, as opposed to their observed, frequencies are considered. Also, although 2 institutions may have observed frequencies that are quite similar, one may be performing significantly better than expected, while the other is performing significantly worse than expected, once patient characteristics are considered. The change in rank order has potential implications for benchmarking as well as economic implications in the context of pay-for-performance.21 In addition, casemix adjustment removes the incentive for hospitals to limit access to patients at greater risk of complications to lessen the frequency of observed morbidity. Limitations of this work should be noted. Patients in this study were from hospitals that are affiliated with academic institutions, and as such may not be representative of patients throughout the country. However, the general characteristics of the patients illustrate that the population is similar across many dimensions (eg, body mass index, multiple gestations) to a more general US obstetric population. Also, the risk-adjustment models incorporate data that, at present, would not be readily available from administrative databases, and thus could not be easily introduced into widespread use. Yet, the variables in the models could be captured from electronic medical records, which are increasing in prevalence and sophistication. In summary, we believe risk adjustment is necessary if obstetrical outcomes are to be compared meaningfully between institutions. However, the intrainstitutional risk-adjusted probabilities

Research

of different obstetric outcomes are poorly correlated and thus performance based on a single outcome cannot be generalized to overall obstetrical performance. Furthermore, use of up to 4 risk-adjusted outcomes did not allow for a summary assessment of overall hospital obstetric quality. These findings underscore the complexity of quality measurement and that the current methods of summarizing hospital performance in obstetrics should be reappraised. ACKNOWLEDGMENTS We thank the subcommittee members who participated in protocol development and coordination between clinical research centers (Cynthia Milluzzi, RN, and Joan Moss, RNC, MSN), protocol/data management and statistical analysis (Elizabeth Thom, PhD and Yuan Zhao, MS), and protocol development and oversight (Brian M. Mercer, MD). In addition to the authors, other members of the Eunice Kennedy Shriver National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network are as follows: Case Western Reserve Universitye MetroHealth Medical Center, Cleveland, OHeB. Mercer, C. Milluzzi, W. Dalton, T. Dotson, P. McDonald, C. Brezine, A. McGrail; Northwestern University, Chicago, ILeG. Mallett, M. Ramos-Brinson, A. Roy, L. Stein, P. Campbell, C. Collins, N. Jackson, M. Dinsmoor (NorthShore University HealthSystem), J. Senka (NorthShore University HealthSystem), K. Paychek (NorthShore University HealthSystem), A. Peaceman; Columbia University, New York, NYeM. Talucci, M. Zylfijaj, Z. Reid (Drexel University), R. Leed (Drexel U), J. Benson (Christiana Hospital), S. Forester (Christiana Hospital), C. Kitto (Christiana Hospital), S. Davis (St Peter’s University Hospital), M. Falk (St Peter’s University Hospital), C. Perez (St Peter’s University Hospital); University of Utah Health Sciences Center, Salt Lake City, UTeK. Hill, A. Sowles, J. Postma (Latter-day Saints Hospital), S. Alexander (Latter-day Saints Hospital), G. Andersen (Latter-day Saints Hospital), V. Scott (McKayDee), V. Morby (McKay-Dee), K. Jolley (Utah Valley Regional Medical Center), J. Miller (Utah Valley Regional Medical Center), B. Berg (Utah Valley Regional Medical Center); University of North Carolina at Chapel Hill, Chapel Hill, NCeK. Dorman, J. Mitchell, E. Kaluta, K. Clark (WakeMed), K. Spicer (WakeMed), S. Timlin (Rex), K. Wilson (Rex); University of Texas Southwestern Medical Center, Dallas, TXeL. Moseley, M. Santillan, J. Price, K. Buentipo, V. Bludau, T. Thomas, L. Fay, C. Melton, J. Kingsbery, R. Benezue; University of Pittsburgh, Pittsburgh, PAeH. Simhan, M. Bickus, D. Fischer, T. Kamon (deceased), D. DeAngelis; Ohio State University, Columbus, OHeC.

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e8

Research

Obstetrics

Latimer, L. Guzzo (St Ann’s), F. Johnson, L. Gerwig (St Ann’s), S. Fyffe, D. Loux (St Ann’s), S. Frantz, D. Cline, S. Wylie, J. Iams; University of Alabama at Birmingham, Birmingham, ALeM. Wallace, A. Northen, J. Grant, C. Colquitt, D. Rouse, W. Andrews; University of Texas Medical Branch, Galveston, TXeJ. Moss, A. Salazar, A. Acosta, G. Hankins; Wayne State University, Detroit, MIeN. Hauff, L. Palmer, P. Lockhart, D. Driscoll, L. Wynn, C. Sudz, D. Dengate, C. Girard, S. Field; Brown University, Providence, RIeP. Breault, F. Smith, N. Annunziata, D. Allard, J. Silva, M. Gamage, J. Hunt, J. Tillinghast, N. Corcoran, M. Jimenez; University of Texas Health Science Center at Houston-Children’s Memorial Hermann Hospital, Houston, TXeF. Ortiz, P. Givens, B. Rech, C. Moran, M. Hutchinson, Z. Spears, C. Carreno, B. Heaps, G. Zamora; Oregon Health and Science University, Portland, OReJ. Seguin, M. Rincon, J. Snyder, C. Farrar, E. Lairson, C. Bonino, W. Smith (Kaiser Permanente), K. Beach (Kaiser Permanente), S. Van Dyke (Kaiser Permanente), S. Butcher (Kaiser Permanente); George Washington University Biostatistics Center, Washington, DCeE. Thom, Y. Zhao, P. McGee, V. Momirova, R. Palugod, B. Reamer, M. Larsen; and Eunice Kennedy Shriver National Institute of Child Health and Human Development, Bethesda, MDeS. Tolivaisa Members of the external advisory committee: the following experts contributed to choice of outcome measures, without compensation: Peter G. Goldschmidt, MD, DrPH, Health Improvement Institute, Bethesda, MD; Jeffrey B. Gould, MD, MPH, Stanford University, Palo Alto, CA; Bill Munier, MD, Agency for Healthcare Research and Quality, Rockville, MD; Elliott K. Main, MD, California Pacific Medical Center, San Francisco, CA; Patrick S. Romano, MD, MPH, University of CaliforniaeDavis, Sacramento, CA; Kimberly Gregory, MD, MPH, Cedars-Sinai Medical Center, Los Angeles, CA; Gary Hankins, MD, University of Texas Medical Branch, Galveston, TX; George A. Macones, MD, Washington University, St Louis, MO; JeanneMarie M. Guise, MD, Oregon Health and Science University, Portland, OR.

www.AJOG.org REFERENCES 1. Chassin MR, Loeb JM, Schmaltz SP, Wachter RM. Accountability measureseusing measurement to promote quality improvement. N Engl J Med 2010;363:683-8. 2. Burstin HR, Conn A, Setnik G, et al. Benchmarking and quality improvement: the Harvard emergency department quality study. Am J Med 1999;107:437-49. 3. Draycott T, Sibanda T, Laxton C, Winter C, Mahmood T, Fox R. Quality improvement demands quality measurement. BJOG 2010;117: 1571-4. 4. Bailit JL. Measuring the quality of inpatient obstetrical care. Obstet Gynecol Surv 2007;62: 207-13. 5. Shroyer AL, Plomondon ME, Grover FL, Edwards FH. The 1996 coronary artery bypass risk model: the Society of Thoracic Surgeons adult cardiac national database. Ann Thorac Surg 1999;67:1205-8. 6. Kozower BD, Sheng S, O’Brien SM, et al. STS database risk models: predictors of mortality and major morbidity for lung cancer resection. Ann Thorac Surg 2010;90:875-83. 7. Novick RJ, Fox SA, Stitt LW, Forbes TL, Steiner S. Direct comparison of risk-adjusted and non-risk-adjusted CUSUM analyses of coronary artery bypass surgery outcomes. J Thorac Cardiovasc Surg 2006;132:386-91. 8. Grobman WA, Feinglass J, Murthy S. Are the Agency for Healthcare Research and Quality obstetric trauma indicators valid measures of hospital safety? Am J Obstet Gynecol 2006;195:868-74. 9. Landy HJ, Laughon SK, Bailit JL, et al. Characteristics associated with severe perineal and cervical lacerations during vaginal delivery. Obstet Gynecol 2011;117:627-35. 10. Ross-Adjie G, McAllister H, Bradshaw S. Graduated compression stockings for the prevention of postoperative venous thromboembolism in obstetric patients: a best practice implementation project. Int J Evid Based Healthc 2012;10:77-81. 11. Dupont C, Deneux-Tharaux C, Touzet S, et al. Clinical audit: a useful tool for reducing

446.e9 American Journal of Obstetrics & Gynecology NOVEMBER 2013

severe postpartum hemorrhages? Int J Qual Health Care 2011;23:583-9. 12. Weed S, Bastek JA, Sammel MD, Beshara M, Hoffman S, Srinivas SK. Comparing postcesarean infectious complication rates using two different skin preparations. Obstet Gynecol 2011;117:1123-9. 13. Tita AT, Landon MB, Spong CY, et al. Timing of elective repeat cesarean delivery at term and neonatal outcomes. N Engl J Med 2009;360:111-20. 14. Lu MC, Fridman M, Korst LM, et al. Variations in the incidence of postpartum hemorrhage across hospitals in California. Matern Child Health J 2005;9:297-306. 15. Skupski DW, Lowenwirt IP, Weinbaum FI, Brodsky D, Danek M, Eglinton GS. Improving hospital systems for the care of women with major obstetric hemorrhage. Obstet Gynecol 2006;107:977-83. 16. Kohavi R. A study of cross-validation and bootstrap for accuracy estimation and model selection. In: Mellish CS, ed. IJCAI’95 proceedings of the 14th international joint conference on artificial intelligence. Vol 2. San Francisco, CA: Morgan Kaufmann Publishers Inc; 1995. 17. Paul P, Pennell M, Lemeshow S. Standardizing the power of the Hosmer-Lemeshow goodness of fit test in large data sets. Stat Med 2013;32:67-80. 18. Ash AS, Shwartz M, Peköz EA. Comparing outcomes across providers. In: Iezzoni L, ed. Risk adjustment for measuring health care outcomes. Chicago, IL: Health Administration Press; 2003. 19. Pronovost PJ, Berenholtz SM, Ngo K, et al. Developing and pilot testing quality indicators in the intensive care unit. J Crit Care 2003;18: 145-55. 20. Shahian DM, O’Brien SM, Filardo G, et al. The Society of Thoracic Surgeons 2008 cardiac surgery risk models: part 3evalve plus coronary artery bypass grafting surgery. Ann Thorac Surg 2009;88:S43-62. 21. Lindenauer PK, Remus D, Roman S, et al. Public reporting and pay for performance in hospital quality improvement. N Engl J Med 2007;356:486-96.

Obstetrics

www.AJOG.org A PPENDIX Statistics Formulae to estimate individual patient outcome probability and expected hospital outcome frequency Using the final multivariable model, expected outcome probabilities were estimated for each patient and averaged over patients within the same hospital to determine the expected outcome frequency for each institution. This was performed for each outcome and its predefined cohort (denominator), using the parameter estimates from the final multivariable model. Patient level outcome frequency can be estimated as

Research


Estimated Individual Patient Outcome Probability log 1  Estimated Individual Patient Outcome Probability b Patient Characteristics þ . þ bb Patient Characteristics ¼ aþb 1

1

n

n

Estimated Individual Patient Outcome Probability 1
 ¼ b b  aþ b 1 Patient Characteristics1 þ.þ b n Patient Characteristicsn 1þe Hence the hospital level outcome frequency is estimated by averaging the individual patient outcome probabilities. Estimated Hospital Outcome Frequency PN 1 Estimated Individual Patient Outcome Probability ¼ N

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e10

Research

Obstetrics

SUPPLEMENTARY FIGURE 1

Model calibration for postpartum hemorrhage

Model calibration for final multivariable model for outcome of postpartum hemorrhage: observed and expected number of patients within each partition of Hosmer-Lemeshow test by presence or absence of outcome. Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

446.e11 American Journal of Obstetrics & Gynecology NOVEMBER 2013

www.AJOG.org

Obstetrics

www.AJOG.org

Research

SUPPLEMENTARY FIGURE 2

Model calibration for peripartum infection

Model calibration for final multivariable model for outcome of peripartum infection: observed and expected number of patients within each partition of Hosmer-Lemeshow test by presence or absence of outcome. Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e12

Research

Obstetrics

SUPPLEMENTARY FIGURE 3

Model calibration for severe perineal laceration at SVD

Model calibration for final multivariable model for outcome of severe perineal laceration at spontaneous vaginal delivery (SVD): observed and expected number of patients within each partition of Hosmer-Lemeshow test by presence or absence of outcome. Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

446.e13 American Journal of Obstetrics & Gynecology NOVEMBER 2013

www.AJOG.org

Obstetrics

www.AJOG.org

Research

SUPPLEMENTARY FIGURE 4

Model calibration for severe perineal laceration at FVD

Model calibration for final multivariable model for outcome of severe perineal laceration at forcepsassisted vaginal delivery (FVD): observed and expected number of patients within each partition of Hosmer-Lemeshow test by presence or absence of outcome. Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e14

Research

Obstetrics

SUPPLEMENTARY FIGURE 5

Model calibration for severe perineal laceration at VVD

Model calibration for final multivariable model for outcome of severe perineal laceration at vacuumassisted vaginal delivery (VVD): observed and expected number of patients within each partition of Hosmer-Lemeshow test by presence or absence of outcome. Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

446.e15 American Journal of Obstetrics & Gynecology NOVEMBER 2013

www.AJOG.org

Obstetrics

www.AJOG.org

Research

SUPPLEMENTARY FIGURE 6

Model calibration for composite neonatal adverse outcome

Model calibration for final multivariable model for composite neonatal adverse outcome: observed and expected number of patients within each partition of Hosmer-Lemeshow test by presence or absence of outcome. Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e16

Research

Obstetrics

www.AJOG.org

SUPPLEMENTARY TABLE 1

Definitions of outcomes, relevant denominators, maternal characteristics, and neonatal characteristics Variables

Categories

Definition

Venous thromboembolism

Yes/no

Occurrence of either of following at any time between hospital admission (including triage) and discharge
Deep venous thrombosis diagnosed by duplex Doppler, excluding septic pelvic thrombosis
Pulmonary embolism diagnosed by CT, spiral CT scan, or ventilationperfusion lung scan

Postpartum hemorrhage

Yes/no

Occurrence of any of following
Estimated blood loss 1500 mL at delivery or immediate postpartum period; actual value (mL) was obtained from medical records
Blood transfusion at any time between hospital admission (including triage) and discharge
Hysterectomy for hemorrhage, placenta accreta, or atony

Peripartum infection

Yes/no

Occurrence of any of following
Chorioamnionitis noted in medical record
Intrapartum antibiotics for treatment of chorioamnionitis or for intrapartum fever without apparent reason other than chorioamnionitis
Postpartum endometritis clinically diagnosed or medical record indicates treatment for endometritis
Postpartum wound infection that required antibiotics (wound cellulitis)
Postpartum wound reopened to treat fluid collection or infection, including wound dehiscence
Laparotomy or unexpected procedure for infection or wound dehiscence

Severe perineal laceration

Yes/no

Occurrence of either of following
Third-degree perineal laceration defined as laceration, either partial or full-thickness, of anal sphincter requiring suture after delivery
Fourth-degree perineal laceration defined as laceration of rectal mucosa requiring suture after delivery

Composite neonatal adverse outcome

Yes/no

Outcomes

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

446.e17 American Journal of Obstetrics & Gynecology NOVEMBER 2013

Occurrence of any of following
Infant stay greater than maternal stay by 3 calendar days
5-min Apgar score <4
Skeletal fracture noted after delivery, excluding clavicle
Facial nerve palsy noted after delivery, defined as injury to facial nerve resulting in partial or complete paralysis of facial muscles
Brachial plexus palsy noted after delivery, defined as palsy of nerves of brachial plexus with partial or complete paralysis affecting upper extremities
Subgaleal hemorrhage
Ventilator support within 24 h of birth on at least 2 d; includes mechanical ventilation; excludes CPAP
Hypoxic ischemic encephalopathy defined as 1 signs in at least 3 of following 6 categories per methods of Shankaran et al1  Level of consciousness  Spontaneous activity  Posture  Tone  Primitive reflexes (suck or Moro)  Autonomic nervous system (pupils, heart rate, or respiration)
Stillbirth after hospital admission
Neonatal death (continued)

Obstetrics

www.AJOG.org

Research

SUPPLEMENTARY TABLE 1

Definitions of outcomes, relevant denominators, maternal characteristics, and neonatal characteristics (continued) Variables

Categories

Definition

Relevant denominators for severe perineal laceration Singleton, spontaneous vaginal delivery with no shoulder dystocia or placenta previa

Yes/no

Singleton, forceps vaginal delivery with no shoulder dystocia or placenta previa

Yes/no

Singleton, vacuum vaginal delivery with no shoulder dystocia or placenta previa

Yes/no


Singleton defined as 1 fetus at delivery and included multiple fetuses where all but 1 fetus died <20 wk of gestation
Spontaneous vaginal delivery defined as vaginal delivery without use of forceps or vacuum
Shoulder dystocia was defined as present if infant required specific delivery maneuvers
See placenta previa definition under maternal characteristics
Singleton defined as 1 fetus at delivery and included multiple fetuses where all but 1 fetus died <20 wk of gestation
Forceps vaginal delivery defined as vaginal delivery with use of forceps
Shoulder dystocia was present if infant required specific delivery maneuvers
See placenta previa definition under maternal characteristics
Singleton defined as 1 fetus at delivery and included multiple fetuses where all but 1 fetus died <20 wk of gestation
Vacuum vaginal delivery defined as vaginal delivery with use of vacuum
Shoulder dystocia was present if infant required specific delivery maneuvers
See placenta previa definition under maternal characteristics

Relevant denominator for composite neonatal adverse outcome Term, nonanomalous singleton infants

Yes/no


Term defined as delivery 37 wk of gestation (see “Gestational age at delivery, wk” for definition)
Nonanomalous defined as no internal or external major congenital physical abnormality known at birth/identified during delivery hospitalization
Singleton defined as 1 fetus at delivery and included multiple fetuses where all but 1 fetus died <20 wk of gestation

Maternal characteristics Age, y

<20 20-24.9


Age in years at admission

25-29.9 30-34.9 35 Race/ethnicity

Non-Hispanic white Non-Hispanic black


As reported in chart

Non-Hispanic Asian Hispanic Other Not documented Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

(continued)

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e18

Research

Obstetrics

www.AJOG.org

SUPPLEMENTARY TABLE 1

Definitions of outcomes, relevant denominators, maternal characteristics, and neonatal characteristics (continued) Variables

Categories

Body mass index at delivery, kg/m2

<25 25-29.9

Definition
Most recent antepartum weight in medical record was recorded
Last height measured in clinic or reported by patient was recorded

30-34.9 35-39.9 40 Cigarette use during pregnancy

Yes/no

Cocaine or methamphetamine use during pregnancy

Yes/no

Insurance status

Uninsured/self-pay Government-assisted


Included tobacco products only
Self-identified
Included all sources in medical record (patient or laboratory test result, including neonatal test if evidence of maternal cocaine use)
Primary source of medical care payment at time of admission

Private Prenatal care

Yes/no

Obstetric history, 4 category variable

Nulliparous Prior vaginal delivery only


If there was documentation of >2 prenatal care visits from any type of care provider, inside or outside hospital system, excluding emergency department or triage visits
Nulliparous defined as no previous pregnancy lasting 20 wk of gestation

Prior cesarean only Prior cesarean and vaginal Obstetric history, 5 category variable

Nulliparous Prior vaginal delivery only Prior cesarean only without classic, T, or J uterine incision


Nulliparous defined as no previous pregnancy lasting 20 wk of gestation
Prior classic, T, or J uterine incision was noted if documented in medical record

Prior cesarean without classic, T, or J uterine incision and vaginal Prior cesarean with classic, T, or J uterine incision (with or without prior vaginal) Prior vaginal delivery

Yes/no

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

446.e19 American Journal of Obstetrics & Gynecology NOVEMBER 2013


Prior vaginal delivery 20 wk of gestation, irrespective of whether or not there was prior cesarean delivery (continued)

Obstetrics

www.AJOG.org

Research

SUPPLEMENTARY TABLE 1

Definitions of outcomes, relevant denominators, maternal characteristics, and neonatal characteristics (continued) Variables

Categories

Any hypertension

Yes/no

Diabetes mellitus

None Gestational

Definition
Included any of following diagnoses documented in medical record:  Chronic hypertension  Gestational hypertension  Preeclampsia  HELLP syndrome (hemolysis, elevated liver enzymes, and low platelet count)  Eclampsia


Diabetes status per diagnosis documented in medical record

Pregestational Anticoagulant use during pregnancy

Yes/no

Multiple gestation

Yes/no

Polyhydramnios

Yes/no

Oligohydramnios

Yes/no

Placenta previa

Yes/no

Placenta accreta

Yes/no

Placental abruption

Yes/no

PROM/PPROM

Yes/no

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.


Included heparin, low-molecular-weight heparin, warfarin and hirudin used during this pregnancy before delivery
2 fetuses at delivery and 2 fetuses survived to 20 wk of gestation
If any of following at time of most recent ultrasound and within 2 wk of hospital admission (including triage):  Diagnosis of polyhydramnios or hydramnios documented in medical record  Amniotic fluid index (AFI) >24 cm  Greatest vertical pocket (GVP) >8 cm


If any of following in nonruptured/intact membranes, at time of most recent ultrasound and within 2 wk of hospital admission (including triage):  Diagnosis of oligohydramnios documented in medical record  Amniotic fluid index (AFI) <5 cm  Greatest vertical pocket (GVP) <2 cm


If either of following:  Hemorrhage diagnosis is indicated in chart and reason is placenta previa  Cesarean delivery and indication is placenta previa


If placenta was adherent to uterine wall without easy separation, at or immediately after delivery
If reason for labor and delivery admission or reason for hospital admission was vaginal bleeding / placental abruption
PROM defined as membrane rupture 12 h before labor and delivery admission
PPROM defined as spontaneous rupture of membranes 1 h before onset of labor (contractions) at 36 wk 6 d of gestation (continued)

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e20

Research

Obstetrics

www.AJOG.org

SUPPLEMENTARY TABLE 1

Definitions of outcomes, relevant denominators, maternal characteristics, and neonatal characteristics (continued) Variables

Categories

GBS status

Negative Positive Unknown

Definition
GBS status per most recent GBS test prior to delivery, excluding throat cultures
Negative included negative vaginal or rectal culture
Positive included positive urine, vaginal, rectal, or unknown source culture or previously affected child
Unknown included no explicit mention in medical record or if results were not available until after delivery

Neonatal characteristics Presentation at delivery

Vertex Breech Nonbreech malpresentation

Gestational age at delivery, wk

230-276 280-336 340-366 370-376


Vertex included compound presentations if it included vertex; included face and brow malpositions
Nonbreech malpresentation included transverse or oblique, or any compound presentation that did not include vertex (eg, only foot and hand)
Based on EDC (estimated delivery date at 40 wk of gestation), which was based on best obstetric estimate on admission note. If actual EDC was not documented, weeks of gestation on specific date was used to calculate EDC

380-386 390-396 400-406 410-416 420 Birthweight, g

<2500 2500-3999


Abstracted from medical record

4000 Size for gestational age

Small Appropriate Large


Estimated per methods of Alexander et al2 using neonatal gestational age at delivery, birthweight and sex and maternal race/ethnicity  Small defined as <10th percentile for sex and race/ethnicity  Appropriate defined as 10th-90th percentile for sex and race/ ethnicity  Large defined as >90th percentile for sex and race/ethnicity

CPAP, continuous positive airway pressure; CT, computed tomography; EDC, estimated date of confinement; GBS, group B streptococcus; PPROM, preterm premature rupture of membranes; PROM, premature rupture of membranes; Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

446.e21 American Journal of Obstetrics & Gynecology NOVEMBER 2013

Obstetrics

www.AJOG.org

Research

SUPPLEMENTARY TABLE 2

Variables assessed in multivariable modeling for each obstetric outcome Postpartum hemorrhage

Peripartum infection

Severe perineal laceration at SVDa

Composite neonatal adverse outcomeb

Age (<20, 20-24.9, 25-29.9, 30-34.9, 35 y)

A, S

A, S

A, S

A, NS

Body mass index at delivery (<25, 25-29.9, 30-34.9, 35-39.9, 40 kg/m2)

A, NS

A, S

A, S

A, S

Cigarette use during pregnancy

A, NS

A, S

A, S

A, S

Cocaine or methamphetamine use during pregnancy

A, NS

NA

NA

A, S

Insurance status (uninsured/self-pay, government-assisted, private)

A, S

A, S

A, S

A, S

Prenatal care

A, S

A, NS

A, NS

A, S

Obstetric history (nulliparous, prior vaginal delivery only, prior cesarean only, prior cesarean, and vaginal)c

A, S

A, S

A, S

A, S

Any hypertension

A, S

A, NS

A, NS

A, S

Diabetes mellitus (none, gestational, pregestational)

A, S

A, S

A, NS

A, S

Anticoagulant use during pregnancy

A, S

NA

NA

NA

Multiple gestation

A, S

A, NS

N/A

N/A

Polyhydramnios

A, NS

NA

NA

NA

Placenta previa

A, S

NA

N/A

A, NS

Placenta accreta

A, S

NA

NA

NA

Placental abruption

A, S

NA

NA

A, NS

PROM/PPROM

A, NS

A, S

NA

A, S

GBS status

A, NS

A, S

NA

A, NS

NA

NA

NA

A, NS

Gestational age at delivery (23 -27 , 28 -33 , 340-366, 370-376, 380-386, 390-396, 400-406, 410-416, 420 wk)

NA

A, S

NA

NA

Birthweight (<2500, 2500-3999, 4000 g)

A, S

NA

A, S

NA

Size for gestational age (small, appropriate, large)

NA

NA

NA

A, S

Variables Maternal characteristics

Neonatal characteristics Presentation at delivery 0

6

0

6

A, assessed in multivariable model; GBS, group B streptococcus; NA, not assessed in multivariable model; N/A, not applicable (excluded from denominator); NS, not significant; PPROM, preterm premature rupture of membranes; PROM, premature rupture of membranes; S, significant; SVD, spontaneous vaginal delivery. a

Among women with singleton delivery and no shoulder dystocia or placenta previa; b Among women with term, nonanomalous singleton infants; c For composite neonatal adverse outcome, obstetric history categories were: nulliparous; prior vaginal delivery only; prior cesarean only without classic, T, or J uterine incision; prior vaginal, and cesarean without classic, T, or J uterine incision; prior cesarean with classic, T, or J uterine incision (with or without prior vaginal); for severe perineal laceration outcomes, obstetric history categories were: no prior vaginal delivery; prior vaginal delivery.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e22

Research

Obstetrics

www.AJOG.org

SUPPLEMENTARY TABLE 3

ORs (95% CIs) for patient characteristics in final multivariable model for postpartum hemorrhagea,b Not accounting for patient clustering within hospital

Accounting for patient clustering within hospital in logistic model (hospital fixed effects)

Accounting for patient clustering within hospital in hierarchical model (hospital random effects)

Variables

Parameter estimates

OR (95% CI)

OR (95% CI)

OR (95% CI)

Intercept

e4.0499

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

0.98 (0.83e1.15)

1.02 (0.86e1.20)

1.02 (0.86e1.20)

MATERNAL CHARACTERISTICS Age, y <20

0

20-24.9

e0.0227

25-29.9

e0.0120

0.99 (0.84e1.17)

1.03 (0.87e1.21)

1.02 (0.86e1.21)

30-34.9

0.0664

1.07 (0.90e1.27)

1.10 (0.92e1.31)

1.10 (0.92e1.31)

35

0.2515

1.29 (1.07e1.55)

1.28 (1.06e1.54)

1.28 (1.06e1.54)

Uninsured/self-pay

0.8589

2.36 (2.06e2.70)

1.35 (1.15e1.58)

1.39 (1.18e1.63)

Government-assisted

0.5678

1.76 (1.59e1.96)

1.35 (1.21e1.51)

1.38 (1.23e1.54)

Private

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

0.62 (0.50e0.78)

0.61 (0.49e0.77)

0.61 (0.49e0.78)

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

e0.5290

0.59 (0.53e0.66)

0.58 (0.52e0.64)

0.58 (0.52e0.64)

Prior cesarean only

0.2420

1.27 (1.10e1.47)

1.22 (1.05e1.41)

1.22 (1.05e1.42)

Prior cesarean and vaginal

0.3511

1.42 (1.24e1.62)

1.39 (1.21e1.59)

1.39 (1.21e1.59)

0.9180

2.50 (2.27e2.77)

2.41 (2.18e2.67)

2.42 (2.17e2.69)

Insurance status

Prenatal care

e0.4728

Obstetric history Nulliparous Prior vaginal delivery only

Any hypertension

0

Diabetes mellitus None

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

Gestational

0.1176

1.13 (0.96e1.31)

1.16 (0.99e1.35)

1.16 (0.99e1.36)

Pregestational

0.3713

1.45 (1.15e1.83)

1.48 (1.17e1.87)

1.48 (1.16e1.88)

Anticoagulant use during pregnancy

0.9471

2.58 (1.92e3.47)

2.45 (1.82e3.31)

2.47 (1.80e3.38)

Multiple gestation

1.3583

3.89 (3.30e4.58)

3.92 (3.32e4.63)

3.92 (3.29e4.66)

Placenta previa

1.7748

5.90 (4.32e8.05)

5.86 (4.28e8.03)

5.86 (4.20e8.16)

Placenta accreta

4.7760

118.63 (76.16e184.77)

131.41 (83.86e205.90)

129.65 (80.79e208.05)

Placental abruption

1.4559

4.29 (3.36e5.48)

4.51 (3.52e5.79)

4.50 (3.47e5.85)

0.3062

1.36 (1.20e1.54)

1.37 (1.21e1.55)

1.37 (1.21e1.56)

NEONATAL CHARACTERISTICS Birthweight, g <2500 2500-3999

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

4000

0.5562

1.74 (1.52e2.01)

1.73 (1.50e1.99)

1.73 (1.50e1.99)

CI, confidence interval; OR, odds ratio; Ref, reference. a

2425 outcomes in denominator size of 105,987; b Among all women with complete outcome and covariable data.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

446.e23 American Journal of Obstetrics & Gynecology NOVEMBER 2013

Obstetrics

www.AJOG.org

Research

SUPPLEMENTARY TABLE 4

ORs (95% CIs) for patient characteristics in final multivariable model for peripartum infectiona,b Not accounting for patient clustering within hospital

Accounting for patient clustering within hospital in logistic model (hospital fixed effects)

Accounting for patient clustering within hospital in hierarchical model (hospital random effects)

Variables

Parameter estimates

OR (95% CI)

OR (95% CI)

OR (95% CI)

Intercept

e3.0409

MATERNAL CHARACTERISTICS Age, y <20

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

20-24.9

0.0692

1.07 (0.98e1.18)

1.07 (0.98e1.18)

1.07 (0.97e1.18)

25-29.9

0.0946

1.10 (1.00e1.21)

1.09 (0.99e1.21)

1.09 (0.99e1.21)

30-34.9

0.0866

1.09 (0.98e1.21)

1.03 (0.92e1.15)

1.03 (0.92e1.15)

0.2056

1.23 (1.09e1.38)

1.14 (1.01e1.29)

1.14 (1.01e1.29)

<25

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

25-29.9

0.1586

1.17 (1.06e1.29)

1.17 (1.06e1.29)

1.17 (1.06e1.29)

35 Body mass index at delivery, kg/m

2

30-34.9

0.3554

1.43 (1.29e1.58)

1.44 (1.31e1.59)

1.44 (1.30e1.60)

35-39.9

0.3457

1.41 (1.26e1.58)

1.47 (1.31e1.65)

1.47 (1.31e1.65)

40

0.3645

1.44 (1.27e1.63)

1.53 (1.35e1.74)

1.53 (1.35e1.74)

e0.2183

0.80 (0.73e0.89)

0.90 (0.82e1.00)

0.90 (0.81e1.01)

Uninsured/self-pay

0.6177

1.86 (1.68e2.04)

1.49 (1.33e1.66)

1.50 (1.34e1.67)

Government-assisted

0.4705

1.60 (1.49e1.72)

1.48 (1.37e1.59)

1.48 (1.37e1.60)

Private

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

0

Cigarette use during pregnancy Insurance status

Obstetric history Nulliparous

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

Prior vaginal delivery only

e1.6803

0.19 (0.17e0.20)

0.18 (0.17e0.20)

0.18 (0.17e0.20)

Prior cesarean only

e1.4300

0.24 (0.21e0.28)

0.24 (0.21e0.28)

0.24 (0.21e0.28)

Prior cesarean and vaginal

e1.2864

0.28 (0.24e0.32)

0.29 (0.25e0.33)

0.29 (0.25e0.33)

Diabetes mellitus None

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

Gestational

0.0986

1.10 (0.98e1.25)

1.15 (1.02e1.30)

1.15 (1.01e1.30)

Pregestational

0.3135

1.37 (1.10e1.70)

1.35 (1.08e1.67)

1.35 (1.08e1.68)

0.8254

2.28 (2.05e2.54)

2.24 (2.01e2.50)

2.24 (2.00e2.52)

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

e0.1700

0.84 (0.78e0.91)

0.85 (0.79e0.92)

0.85 (0.79e0.92)

0.1413

1.15 (1.06e1.25)

0.71 (0.63e0.79)

0.71 (0.64e0.80)

PROM/PPROM GBS status Negative Positive Unknown

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

(continued)

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e24

Research

Obstetrics

www.AJOG.org

SUPPLEMENTARY TABLE 4

ORs (95% CIs) for patient characteristics in final multivariable model for peripartum infectiona,b (continued)

Variables

Not accounting for patient clustering within hospital

Accounting for patient clustering within hospital in logistic model (hospital fixed effects)

Accounting for patient clustering within hospital in hierarchical model (hospital random effects)

Parameter estimates

OR (95% CI)

OR (95% CI)

OR (95% CI)

NEONATAL CHARACTERISTICS Gestational age at delivery, wk 230-276

1.2949

3.65 (3.04e4.38)

5.04 (4.18e6.09)

5.00 (4.14e6.05)

0

28 -33

6

0.4436

1.56 (1.35e1.80)

2.05 (1.76e2.38)

2.04 (1.75e2.37)

0

34 -36

6

e0.5883

0.56 (0.48e0.64)

0.68 (0.58e0.78)

0.67 (0.58e0.78)

0

37 -37

6

e0.1767

0.84 (0.74e0.95)

0.87 (0.77e0.99)

0.87 (0.77e0.99)

0

6

e0.0951

38 -38

390-396

0.91 (0.83e1.00)

0.92 (0.84e1.01)

0.92 (0.84e1.01)

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

0

40 -40

6

0.3399

1.41 (1.30e1.52)

1.39 (1.29e1.50)

1.39 (1.29e1.50)

0

6

0.6050

1.83 (1.67e2.01)

1.76 (1.60e1.93)

1.76 (1.60e1.93)

0.7091

2.03 (1.55e2.67)

1.70 (1.29e2.24)

1.70 (1.29e2.25)

41 -41 42

0

CI, confidence interval; GBS, group B streptococcus; OR, odds ratio; PPROM, preterm premature rupture of membranes; PROM, premature rupture of membranes; Ref, reference. a

5581 outcomes in denominator size of 110,205; b Among all women with complete outcome and covariable data.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

446.e25 American Journal of Obstetrics & Gynecology NOVEMBER 2013

Obstetrics

www.AJOG.org

Research

SUPPLEMENTARY TABLE 5

ORs (95% CIs) for patient characteristics in final multivariable model for severe perineal laceration at spontaneous vaginal deliverya,b Not accounting for patient clustering within hospital

Accounting for patient clustering within hospital in logistic model (hospital fixed effects)

Accounting for patient clustering within hospital in hierarchical model (hospital random effects)

Variables

Parameter estimates

OR (95% CI)

OR (95% CI)

OR (95% CI)

Intercept

e3.3261

MATERNAL CHARACTERISTICS Age, y <20

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

20-24.9

0.1478

1.16 (0.94e1.43)

1.17 (0.95e1.45)

1.17 (0.94e1.45)

25-29.9

0.6648

1.94 (1.58e2.39)

1.99 (1.62e2.46)

1.99 (1.61e2.45)

30-34.9

0.7795

2.18 (1.76e2.71)

2.42 (1.94e3.01)

2.39 (1.91e2.98)

0.6980

2.01 (1.57e2.58)

2.31 (1.79e2.98)

2.27 (1.76e2.94)

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

0.95 (0.81e1.10)

0.94 (0.80e1.09)

0.94 (0.80e1.10)

35 Body mass index at delivery, kg/m

2

<25 25-29.9

e0.0561

30-34.9

e0.1483

0.86 (0.73e1.02)

0.84 (0.71e0.99)

0.84 (0.71e1.00)

35-39.9

e0.2879

0.75 (0.60e0.93)

0.72 (0.58e0.90)

0.72 (0.58e0.90)

40

e0.3060

0.74 (0.56e0.97)

0.69 (0.52e0.91)

0.70 (0.53e0.92)

e0.6517

0.52 (0.39e0.69)

0.49 (0.37e0.66)

0.50 (0.37e0.67)

0.1326

1.14 (0.95e1.37)

1.13 (0.91e1.40)

1.12 (0.90e1.39)

e0.2726

0.76 (0.66e0.88)

0.79 (0.67e0.92)

0.78 (0.67e0.91)

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

e2.2125

0.11 (0.10e0.13)

0.10 (0.09e0.12)

0.11 (0.09e0.12)

e1.6606

0.19 (0.12e0.30)

0.19 (0.12e0.30)

0.19 (0.12e0.31)

Cigarette use during pregnancy Insurance status Uninsured/self-pay Government-assisted Private Prior vaginal delivery

0

NEONATAL CHARACTERISTICS Birthweight, g <2500 2500-3999

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

4000

1.1657

3.21 (2.74e3.75)

3.20 (2.74e3.75)

3.21 (2.73e3.77)

CI, confidence interval; OR, odds ratio; Ref, reference. a

1475 outcomes in denominator size of 68,144; b Among women with singleton delivery and no shoulder dystocia or placenta previa and complete outcome and covariable data.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e26

Research

Obstetrics

www.AJOG.org

SUPPLEMENTARY TABLE 6

ORs (95% CIs) for patient characteristics in final multivariable model for severe perineal laceration at forceps-assisted vaginal deliverya,b Not accounting for patient clustering within hospital

Accounting for patient clustering within hospital in logistic model (hospital fixed effects)

Accounting for patient clustering within hospital in hierarchical model (hospital random effects)

Variables

Parameter estimates

OR (95% CI)

OR (95% CI)

OR (95% CI)

Intercept

e0.6376

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

0.64 (0.42e0.97)

0.81 (0.52e1.25)

0.78 (0.50e1.20)

MATERNAL CHARACTERISTICS Age, y <20

0

20-24.9

e0.4443

25-29.9

e0.1872

0.83 (0.54e1.26)

0.83 (0.54e1.30)

0.83 (0.54e1.29)

30-34.9

0.3339

1.40 (0.92e2.13)

1.17 (0.75e1.84)

1.20 (0.76e1.87)

0.0740

1.08 (0.67e1.74)

0.81 (0.49e1.34)

0.85 (0.51e1.40)

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

0.97 (0.72e1.30)

0.94 (0.69e1.28)

0.94 (0.69e1.28)

35 Body mass index at delivery, kg/m

2

<25 25-29.9

e0.0348

30-34.9

0.0239

1.02 (0.74e1.43)

0.97 (0.69e1.36)

0.98 (0.70e1.38)

35-39.9

e0.4334

0.65 (0.40e1.06)

0.61 (0.37e1.02)

0.63 (0.38e1.05)

40

e0.1131

0.89 (0.52e1.55)

0.94 (0.53e1.66)

0.90 (0.51e1.59)

e0.5771

0.56 (0.33e0.96)

0.63 (0.36e1.12)

0.61 (0.34e1.10)

0.2707

1.31 (0.91e1.89)

0.98 (0.62e1.53)

1.02 (0.65e1.58)

e0.0832

0.92 (0.68e1.24)

0.82 (0.58e1.15)

0.83 (0.59e1.16)

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

e1.4395

0.24 (0.17e0.33)

0.26 (0.18e0.37)

0.26 (0.18e0.38)

e0.3640

0.70 (0.41e1.17)

0.69 (0.40e1.18)

0.69 (0.40e1.20)

Cigarette use during pregnancy Insurance status Uninsured/self-pay Government-assisted Private Prior vaginal delivery

0

NEONATAL CHARACTERISTICS Birthweight, g <2500 2500-3999

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

4000

0.6086

1.84 (1.23e2.74)

1.98 (1.31e3.01)

1.96 (1.28e2.99)

CI, confidence interval; OR, odds ratio; Ref, reference. a

523 outcomes in denominator size of 1898; b Among women with singleton delivery and no shoulder dystocia or placenta previa and complete outcome and covariable data.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

446.e27 American Journal of Obstetrics & Gynecology NOVEMBER 2013

Obstetrics

www.AJOG.org

Research

SUPPLEMENTARY TABLE 7

ORs (95% CIs) for patient characteristics in final multivariable model for severe perineal laceration at vacuum-assisted vaginal deliverya,b Not accounting for patient clustering within hospital

Accounting for patient clustering within hospital in logistic model (hospital fixed effects)

Accounting for patient clustering within hospital in hierarchical model (hospital random effects)

Variables

Parameter estimates

OR (95% CI)

OR (95% CI)

OR (95% CI)

Intercept

e1.7225

MATERNAL CHARACTERISTICS Age, y <20

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

20-24.9

0.3532

1.42 (0.97e2.10)

1.37 (0.92e2.04)

1.40 (0.94e2.08)

25-29.9

0.6489

1.91 (1.30e2.83)

1.76 (1.18e2.62)

1.83 (1.22e2.73)

30-34.9

0.4695

1.60 (1.06e2.41)

1.48 (0.97e2.26)

1.52 (0.99e2.31)

0.4461

1.56 (0.99e2.48)

1.38 (0.86e2.22)

1.46 (0.91e2.35)

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

0.93 (0.71e1.20)

0.95 (0.72e1.23)

0.94 (0.72e1.23)

35 Body mass index at delivery, kg/m

2

<25 25-29.9

e0.0768

30-34.9

e0.2944

0.75 (0.55e1.00)

0.75 (0.55e1.01)

0.75 (0.56e1.02)

35-39.9

e0.7865

0.46 (0.29e0.73)

0.44 (0.27e0.70)

0.45 (0.28e0.72)

40

e0.4673

0.63 (0.37e1.08)

0.60 (0.35e1.05)

0.62 (0.36e1.08)

e0.4777

0.62 (0.40e0.96)

0.58 (0.37e0.90)

0.61 (0.38e0.97)

0.2935

1.34 (0.90e1.99)

1.38 (0.90e2.13)

1.35 (0.87e2.07)

e0.1249

0.88 (0.68e1.15)

0.97 (0.73e1.29)

0.92 (0.69e1.22)

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

e1.4863

0.23 (0.17e0.31)

0.24 (0.18e0.33)

0.24 (0.17e0.33)

e1.2024

0.30 (0.14e0.65)

0.32 (0.15e0.69)

0.31 (0.14e0.69)

Cigarette use during pregnancy Insurance status Uninsured/self-pay Government-assisted Private Prior vaginal delivery

0

NEONATAL CHARACTERISTICS Birthweight, g <2500 2500-3999

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

4000

0.9071

2.48 (1.74e3.54)

2.61 (1.82e3.75)

2.54 (1.76e3.68)

CI, confidence interval; OR, odds ratio; Ref, reference. a

510 outcomes in denominator size of 3515; b Among women with singleton delivery and no shoulder dystocia or placenta previa and complete outcome and covariable data.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e28

Research

Obstetrics

www.AJOG.org

SUPPLEMENTARY TABLE 8

ORs (95% CIs) for patient characteristics in final multivariable model for composite neonatal adverse outcomea,b Not accounting for patient clustering within hospital

Accounting for patient clustering within hospital in logistic model (hospital fixed effects)

Accounting for patient clustering within hospital in hierarchical model (hospital random effects)

Variables

Parameter estimates

OR (95% CI)

OR (95% CI)

Intercept

e3.5780

OR (95% CI)

MATERNAL CHARACTERISTICS Body mass index at delivery, kg/m2 <25

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

25-29.9

0.0513

1.05 (0.91e1.22)

1.07 (0.92e1.23)

1.06 (0.92e1.23)

30-34.9

0.1433

1.15 (1.00e1.34)

1.17 (1.01e1.36)

1.17 (1.01e1.36)

35-39.9

0.2377

1.27 (1.08e1.50)

1.28 (1.08e1.51)

1.28 (1.08e1.51)

40

0.3846

1.47 (1.23e1.75)

1.47 (1.23e1.75)

1.47 (1.23e1.75)

Cigarette use during pregnancy

0.7414

2.10 (1.88e2.34)

2.05 (1.83e2.29)

2.05 (1.82e2.30)

Cocaine or methamphetamine use during pregnancy

1.6714

5.32 (4.21e6.72)

5.53 (4.37e7.00)

5.51 (4.30e7.06)

Uninsured/self-pay

0.2275

1.26 (1.08e1.45)

1.32 (1.12e1.56)

1.33 (1.12e1.57)

Government-assisted

0.4439

1.56 (1.42e1.71)

1.50 (1.35e1.65)

1.51 (1.36e1.67)

Private

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

0.58 (0.46e0.74)

0.62 (0.48e0.79)

0.61 (0.48e0.79)

Insurance status

Prenatal care

e0.5380

Obstetric history 1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

Prior vaginal delivery only

Nulliparous

e0.3415

0.71 (0.65e0.78)

0.71 (0.65e0.78)

0.71 (0.65e0.78)

Prior cesarean only without classic, T, or J uterine incision

e0.2698

0.76 (0.64e0.91)

0.78 (0.66e0.93)

0.78 (0.65e0.93)

Prior cesarean without classic, T, or J uterine incision and vaginal

e0.4095

0.66 (0.56e0.79)

0.67 (0.56e0.79)

0.67 (0.56e0.79)

Prior cesarean with classic, T, or J uterine incision (with or without prior vaginal)

e0.4180

0.66 (0.43e1.01)

0.64 (0.41e0.98)

0.64 (0.41e0.99)

0.3397

1.41 (1.25e1.58)

1.42 (1.26e1.60)

1.42 (1.25e1.61)

None

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

Gestational

0.2003

1.22 (1.04e1.44)

1.26 (1.07e1.48)

1.26 (1.06e1.49)

Pregestational

0.8621

2.37 (1.85e3.03)

2.40 (1.87e3.07)

2.40 (1.86e3.09)

0.4808

1.62 (1.25e2.09)

1.62 (1.25e2.09)

1.62 (1.23e2.12)

Small

0.8391

2.31 (2.07e2.59)

2.29 (2.04e2.56)

2.29 (2.04e2.57)

Appropriate

0

1.00 (Ref)

1.00 (Ref)

1.00 (Ref)

Large

0.4499

1.57 (1.37e1.79)

1.61 (1.41e1.83)

1.60 (1.40e1.83)

Any hypertension

0

Diabetes mellitus

PROM/PPROM NEONATAL CHARACTERISTICS Size for gestational age

CI, confidence interval; OR, odds ratio; PPROM, preterm premature rupture of membranes; PROM, premature rupture of membranes; Ref, reference. a

2440 outcomes in denominator size of 89,279; b Among women with term, nonanomalous singleton infants and complete outcome and covariable data.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

446.e29 American Journal of Obstetrics & Gynecology NOVEMBER 2013

Research

Obstetrics

www.AJOG.org

SUPPLEMENTARY TABLE 9

Concordance between unadjusted and adjusted hospital ranks

Variable

Denominator size for each outcome

Range of absolute rank difference between unadjusted and adjusted rank

Postpartum hemorrhage

105,987

0e6

1

0.97

Peripartum infection

Kendell coefficient of concordancea

110,205

0e5

1

0.98

b

68,144

0e12

3

0.86

b

1898

0e8

2

0.96

3515

0e5

1.5

0.97

0e12

2

0.90

Severe perineal laceration at SVD Severe perineal laceration at FVD

Median absolute rank difference between unadjusted and adjusted rank

Severe perineal laceration at VVD

b

Composite neonatal adverse outcomec

89,279

FVD, forceps-assisted vaginal delivery; SVD, spontaneous vaginal delivery; VVD, vacuum-assisted vaginal delivery. a

P < .001 for all; b Among women with singleton delivery and no shoulder dystocia or placenta previa; c Among women with term, nonanomalous singleton infants.

Bailit. Variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013.

SUPPLEMENTARY REFERENCES 1. Shankaran S, Laptook AR, Ehrenkranz RA, et al. National Institute of Child Health and

Human Development Neonatal Research Network Whole-body hypothermia for neonates with hypoxic-ischemic encephalopathy. N Engl J Med 2005;353:1574-84.

2. Alexander GR, Kogan MD, Himes JH. 19941996 US singleton birth weight percentiles for gestational age by race, Hispanic origin, and gender. Matern Child Health J 1999;3:225-31.

NOVEMBER 2013 American Journal of Obstetrics & Gynecology

446.e30