Академический Документы
Профессиональный Документы
Культура Документы
org
OBSTETRICS
Prediction of cesarean delivery in the term nulliparous
woman: results from the prospective, multicenter
Genesis study
Naomi Burke, MRCOG; Gerard Burke, MD; Fionnuala Breathnach, MD; Fionnuala McAuliffe, MD; John J. Morrison, MD;
Michael Turner, MD; Samina Dornan, MD; John R. Higgins, MD; Amanda Cotter, MD; Michael Geary, MD;
Peter McParland, MD; Sean Daly, MD; Fiona Cody, MSc; Pat Dicker, PhD; Elizabeth Tully, PhD;
Fergal D. Malone, MD; on behalf of the Perinatal Ireland Research Consortium
BACKGROUND: In contemporary practice many nulliparous women RESULTS: From a total enrolled cohort of 2336 nulliparous participants,
require intervention during childbirth such as operative vaginal delivery or 491 (21%) had an unplanned CD. Five parameters were determined to be
cesarean delivery (CD). Despite the knowledge that the increasing rate of the best combined predictors of CD. These were advancing maternal age
CD is associated with increasing maternal age, obesity and larger infant (odds ratio [OR], 1.21; 95% confidence interval [CI], 1.09 to 1.34), shorter
birthweight, we lack a reliable method to predict the requirement for such maternal height (OR, 1.72; 95% CI, 1.52 to 1.93), increasing body mass
potentially hazardous obstetric procedures during labor and delivery. This index (OR, 1.29; 95% CI, 1.17 to 1.43), larger fetal abdominal circum-
issue is important, as there are greater rates of morbidity and mortality ference (OR, 1.23; 95% CI, 1.1 to 1.38), and larger fetal head circum-
associated with unplanned CD performed in labor compared with ference (OR, 1.27; 95% CI, 1.14 to 1.42). A nomogram was developed to
scheduled CDs. A prediction algorithm to identify women at risk of an provide an individualized risk assessment to predict CD in clinical practice,
unplanned CD could help reduced labor associated morbidity. with excellent calibration and discriminative ability (KolmogoroveSmirnov,
OBJECTIVE: In this primary analysis of the Genesis study, our objective D statistic, 0.29; 95% CI, 0.28 to 0.30) with a misclassification rate of 0.21
was to prospectively assess the use of prenatally determined, maternal (95% CI, 0.19 to 0.25).
and fetal, anthropomorphic, clinical, and ultrasound features to develop a CONCLUSION: Five parameters (maternal age, body mass index,
predictive tool for unplanned CD in the term nulliparous woman, before the height, fetal abdominal circumference, and fetal head circumference)
onset of labor. can, in combination, be used to better determine the overall risk
MATERIALS AND METHODS: The Genesis study recruited 2336 of CD in nulliparous women at term. A risk score can be used to
nulliparous women with a vertex presentation between 39þ0 and 40þ6 inform women of their individualized probability of CD. This risk tool
weeks’ gestation in a prospective multicenter national study to examine may be useful for reassuring most women regarding their likely
predictors of CD. At recruitment, a detailed clinical evaluation and ultra- success at achieving an uncomplicated vaginal delivery as well
sound assessment were performed. To reduce bias from knowledge of as selecting those patients with such a high risk for CD that they
these data potentially influencing mode of delivery, women, midwives, and should avoid a trial of labor. Such a risk tool has the potential to
obstetricians were blinded to the ultrasound data. All hypothetical prenatal greatly improve planning hospital service needs and minimizing
risk factors for unplanned CD were assessed as a composite. Multiple patient risk.
logistic regression analysis and mathematical modeling was used to
develop a risk evaluation tool for CD in nulliparous women. Continuous Key words: cesarean delivery, Genesis study, nulliparous women,
predictors were standardized using z scores. prediction algorithm, predictive tool
deliveries
and 40þ6 weeks’ gestation. Baseline
characteristics were obtained such as Failure to progress 504 37.2%
age, weight, height, body mass index CTG abnormalities 778 57.4%
(BMI) (all obtained at the first antenatal Abnormal fetal blood sample 85 6.3%
visit), gestational weight gain assessed at
Suspected infection or sepsis 88 6.5%
the study visit, whether they had been
screened for GDM and the screening Malposition 41 3.0%
method, maternal head circumference Other 53 3.9%
(HC), ethnicity, attendance at prenatal Postpartum hemorrhage (estimated and measured blood 279 11.9%
classes, model of prenatal care, presence loss over 500 mL)
of written birth plan, medical history, Anal sphincter injury 76 3.2%
previous history of cervical surgery,
Shoulder dystocia 29 1.2%
family history of CD in a first-degree
relative, marital status, smoking status, NICU admission 177 7.6%
alcohol and drug use, employment de- CTG, cardiotocography; NICU, neonatal intensive care unit.
tails, and highest level of education a
Suspected maternal compromise included; hypertension, vaginal bleeding, abdominal pain, exhaustion, discomfort; b Sus-
pected fetal compromise included; suspicious CTG pattern, reduced fetal movements, abnormal biophysical score; c Indi-
achieved. cation for delivery: more than 1 indication for delivery may have been recorded.
A study ultrasound examination was Burke et al. Prediction of cesarean delivery in nulliparas. Am J Obstet Gynecol 2017.
performed after 39 completed weeks and
before 40 weeks and 6 days. Standard
fetal biometry was measured including formula. The fetal head position and Managing clinicians and midwives also
biparietal diameter, HC, abdominal engagement were recorded at the time of were blinded to the results of fetal
circumference (AC), and femur length, the study ultrasound. Biometric data biometry, so as to avoid any potential for
yielding a calculated estimated fetal from this ultrasound examination were bias from suspicion of fetal macrosomia
weight (EFW) via the Hadlock-4 not revealed to study participants. influencing decisions relating to timing
TABLE 2
Best Combined predictors of cesarean delivery
At initial prenatal visit At 39þ0 to 40þ6 weeks
Demographic/ultrasound information Mean SD OR unit OR (95% CI) OR (95% CI)
Age, y 29.9 5.07 þ1 SD 1.22 (1.10,1.35) 1.21 (1.09,1.34)
Height, cm 165.5 6.55 1 SD 1.59 (1.43,1.78) 1.72 (1.52,1.93)
BMI, kg/m 2
24.5 4.27 þ1 SD 1.32 (1.20,1.46) 1.29 (1.17,1.43)
Fetal HC, mm 337 12.9 þ1 SD Not applicable 1.27 (1.14,1.42)
Fetal AC, mm 351 16 þ1 SD Not applicable 1.23 (1.10,1.38)
Results from a multiple logistic regression of z-scores are displayed. OR correspond to a þ/1 SD increase/reduction in a predictor. This table can be used to calculate z scores.
AC, abdominal circumference; BMI, body mass index; CI, confidence interval; HC, head circumference; OR, odds ratio; SD, standard deviation.
Burke et al. Prediction of cesarean delivery in nulliparas. Am J Obstet Gynecol 2017.
and mode of delivery. A biophysical error, the overall calibration, and inter- statistical modeling. Stata version 13 and
score was performed and documented in nal validation of the prediction model. the nomolog package was used for the
the participants medical records. Each of these assessments is outlined in nomogram construction.19
Findings of an abnormal biophysical detail here.
profile (<6/8), a diagnosis of small-for- The Akaike information criterion Results
gestational age, or an EFW >5000 g (AIC) was used to determine model fit. The study population profile is described
were revealed to the study participants The AIC is the asymptotically equivalent in Figure 1. A total of 2392 study par-
and managing clinicians, with all such to leave-one out cross-validation for ticipants were recruited. There were 56
revealed cases being excluded from the penalizing complexity.15 Validation an- (2.3%) participants excluded for the
study. Standard perinatal and obstetric alyses were performed on the full data following reasons; lost to follow-up (n ¼
data were collected contemporaneously (“apparent” model fit) in addition to 4), abnormal biophysical profile (n ¼
and included gestational age at delivery, resampling methods: full boostrap 33), EFW < 2.5 kg or > 5 kg (n ¼ 5),
onset of labor, use of prostaglandin for sampling with replacement (1000 re- EFW performed after enrollment (n ¼
preinduction cervical priming, amniot- peats) and 10-fold cross-validation (100 5), and preexisting indication for CD,
omy, use of oxytocin, maternal fever, repeats). The Brier score, scaled Brier such as breech presentation and
type of analgesia used, duration of labor, score, and Pietra index were considered placental abruption (n ¼ 9). Therefore, a
mode of delivery, indication for opera- as overall measures of prediction error.16 total of 2336 women were included in
tive delivery, perineal trauma, and blood The Hosmer-Lemeshow test and Stukel the final analysis, and they represented
loss. tests were used to determine adequacy of the cohort of nulliparas who underwent
calibration.17 In addition, calibration a blinded ultrasound evaluation of fetal
Statistical analysis intercept (“calibration-in-the-large”) weight after 39 weeks’ gestational age
The primary objective of the statistical and calibration slope were assessed for and who were deemed suitable for trial
analysis was to predict individual abso- potential model recalibration and pre- of labor at the time of the final ultra-
lute risk of CD with optimal calibration sented using smoothed loess curves.18 sound assessment.
and discriminative ability. Simple and Discriminative ability was assessed The average maternal age was 29 5.1
multiple logistic regression analyses were using the KolmogoroveSmirnov (KS) D years and the average maternal BMI was
used to model the maternal de- statistic, Gini coefficient, area under the 24.5 4.3 kg/m2. Mean maternal
mographics and ultrasound biometry curve c-statistic, and the misclassifica- gestational weight gain was 13.8 kg 5.3
risk factors for CD. Continuous pre- tion rate. Different link functions and kg. The majority of those enrolled were
dictors were standardized using z scores their effect upon calibration were of European ethnicity (n ¼ 2221;
(predictor value e mean/SD) to show assessed, including the complementary 95.2%). More than one half (1215; 52%)
the relative effects of predictors. Pairwise log-log and generalized logistic function. attended for obstetrician-provided care,
interaction terms and quadratic effects Center effects were assessed with the use with 1120 (48%) attending midwifery-
of variables were included in the set of of stratified logistic regression and provided services. In relation to prepa-
potential predictors. In an analysis of a random-coefficients logistic regression. ration for labor, 1807 (77.4%) attended
prediction model, there are several areas SAS Version 9.2 (SAS Institute, Inc, Cary, prenatal education classes with 940
of importance to be considered; the NC) was used for data management and (40.3%) preparing a formal written birth
model performance in terms of the ac- data screening. SAS PROC LOGISTIC plan. The majority of the participants
curacy of predictions, the prediction and PROC NLMIXED were used for had never smoked (1355; 58%), 774
reporting area, 2008. Natl Vital Stat Rep 13. Healthcare Pricing Office. National Perinatal 27. Dudley NJ. A systematic review of the ul-
2011;59:1-13, 16. Statistics Report Ireland. Dublin: Health Service trasound estimation of fetal weight. Ultrasound
2. Gibbons L, Belizan JM, Lauer JA, Betran AP, Executive; 2013. Obstet Gynecol 2005;25:80-9.
Merialdi M, Althabe F. Inequities in the use of 14. Ender P. University Statistical Reporting 28. Little SE, Edlow AG, Thomas AM, Smith NA.
cesarean section deliveries in the world. Am J Group. Powerlog: command for Logistic Estimated fetal weight by ultrasound: a modifi-
Obstet Gynecol 2012;206:331.e1-19. Regression Power Analysis. College Station, TX: able risk factor for cesarean delivery? Am J
3. Allen VM, O’Connell CM, Liston RM, StataCorp; 2006. Obstet Gynecol 2012;207:309.e1-6.
Baskett TF. Maternal morbidity associated with 15. Stone M. An asymptotic equivalence of 29. Grobman WA, Lai Y, Landon MB, et al.
cesarean delivery without labor compared with choice of model by cross-validation and Development of a nomogram for prediction of
spontaneous onset of labor at term. Obstet Akaike’s criterion. J Roy Stat Soc Ser B (Meth- vaginal birth after cesarean delivery. Obstet
Gynecol 2003;102:477-82. odological) 1977;39:44-7. Gynecol 2007;109:806-12.
4. Allen VM, O’Connell CM, Baskett TF. Maternal 16. Wu YC, Lee WC. Alternative performance 30. Mazouni C, Rouzier R, Collette E, et al.
and perinatal morbidity of caesarean delivery at measures for prediction models. PLoS One Development and validation of a nomogram to
full cervical dilatation compared with caesarean 2014;9:e91249. predict the risk of cesarean delivery in macro-
delivery in the first stage of labour. BJOG 17. Hosmer DW, Hosmer T, Le Cessie S, somia. Acta Obstet Gynecol Scand 2008;87:
2005;112:986-90. Lemeshow S. A comparison of goodness-of-fit 518-23.
5. Dietz HP, Campbell S. Toward normal birth- tests for the logistic regression model. Stat 31. Sinnott SJ, Layte R, Brick A, Turner MJ.
but at what cost? Am J Obstet Gynecol Med 1997;16:965-80. Variation in induction of labour rates across Irish
2016;215:439-44. 18. Austin PC, Steyerberg EW. Graphical hospitals; a cross-sectional study. Eur J Public
6. Patel RR, Peters TJ, Murphy DJ, Team AS. assessment of internal and external calibration of Health 2016;26:753-60.
Prenatal risk factors for Caesarean section. logistic regression models by using loess 32. Kleinrouweler CE, Cheong-See FM,
Analyses of the ALSPAC cohort of 12 944 smoothers. Stat Med 2014;33:517-35. Collins GS, et al. Prognostic models in obstet-
women in England. Int J Epidemiol 2005;34: 19. Zlotnik A, Abraira V. A general purpose rics: available, but far from applicable. Am J
353-67. nomogram generator for predictive logistic Obstet Gynecol 2016;214:79-90.e36.
7. American College of Obstetricians and regression. Stata J 2015;15:537-46.
Gynecologists (College); Society for Maternal- 20. Joseph KS, Young DC, Dodds L, et al.
Fetal Medicine, Caughey AB, Cahill AG, Author and article information
Changes in maternal characteristics and obstetric
Guise J-M, Rouse DJ. Safe prevention of the practice and recent increases in primary cesarean From the Royal College of Surgeons in Ireland, Rotunda
primary cesarean delivery. Am J Obstet Gynecol delivery. Obstet Gynecol 2003;102:791-800. Hospital, Dublin, Ireland (Drs Burke, Breathnach, Tully,
2014;210:179-93. 21. Smith GC, Cordeaux Y, White IR, et al. The and Malone and Ms Cody); Department of Obstetrics and
8. Zaretsky MV, Alexander JM, McIntire DD, effect of delaying childbirth on primary cesarean Gynecology, Graduate Entry Medical School, University of
Hatab MR, Twickler DM, Leveno KJ. Magnetic section rates. PLoS Med 2008;5:e144. Limerick, Limerick, Ireland (Drs Burke and Cotter); UCD
resonance imaging pelvimetry and the predic- 22. Poobalan AS, Aucott LS, Gurung T, School of Medicine and Medical Science, National Ma-
tion of labor dystocia. Obstet Gynecol Smith WC, Bhattacharya S. Obesity as an ternity Hospital, Dublin, Ireland (Dr McAuliffe); National
2005;106:919-26. independent risk factor for elective and University of Ireland, Galway, Ireland (Dr Morrison); UCD
9. Pattinson RC. Pelvimetry for fetal cephalic emergency caesarean delivery in nulliparous Center for Human Reproduction Coombe Women and
presentations at term. Cochrane Database Syst women—systematic review and meta-analysis of Infants University Hospital, Dublin, Ireland (Dr Turner);
Rev 2000;(2):CD000161. cohort studies. Obes Rev 2009;10:28-35. Royal Jubilee Maternity Hospital, Belfast, Ireland (Dr
10. Rozenholc AT, Ako SN, Leke RJ, 23. Jolly MC, Sebire NJ, Harris JP, Regan L, Dornan); University College Cork, Cork University Mater-
Boulvain M. The diagnostic accuracy of external Robinson S. Risk factors for macrosomia and its nity Hospital, Cork, Ireland (Dr Higgins); Obstetrics &
pelvimetry and maternal height to predict clinical consequences: a study of 350,311 Gynecology, St. Michael’s Hospital, Toronto, University of
dystocia in nulliparous women: a study in pregnancies. Eur J Obstet Gynecol Reprod Biol Toronto, Toronto, Canada (Dr Geary); National Maternity
Cameroon. BJOG 2007;114:630-5. 2003;111:9-14. Hospital, Dublin, Ireland (Dr McParland); Coombe Women
11. Al Housseini A, Newman T, Cox A, 24. Kennelly MM, Anjum R, Lyons S, Burke G. and Infants University Hospital, Dublin, Ireland (Drs Daly
Devoe LD. Prediction of risk for cesarean Postpartum fetal head circumference and its and Dicker); and Epidemiology & Public Health, Royal
delivery in term nulliparas: a comparison of influence on labour duration in nullipara. J Obstet College of Surgeons in Ireland, Dublin, Ireland (Dr Dicker).
neural network and multiple logistic regression Gynaecol 2003;23:496-9. Received Nov. 25, 2016; revised Feb. 4, 2017;
models. Am J Obstet Gynecol 2009;201:113. 25. Turner MJ, Rasmussen MJ, Turner JE, accepted Feb. 8, 2017.
e1-6. Boylan PC, MacDonald D, Stronge JM. The in- The authors report no conflict of interest.
12. Eggebo TM, Wilhelm-Benartzi C, fluence of birth weight on labor in nulliparas. This research was funded by the Health Research
Hassan WA, Usman S, Salvesen KA, Lees CC. Obstet Gynecol 1990;76:159-63. Board of Ireland.
A model to predict vaginal delivery in nulliparous 26. Chauhan SP, Grobman WA, Gherman RA, Presented orally at the Society of Maternal Fetal
women based on maternal characteristics and et al. Suspicion and treatment of the macro- Medicine, Atlanta, GA, Feb. 14, 2016.
intrapartum ultrasound. Am J Obstet Gynecol somic fetus: a review. Am J Obstet Gynecol Corresponding author: Naomi Burke, MRCPI, MRCOG.
2015;213:362.e1-6. 2005;193:332-46. naomiburke@rcsi.ie
Glossary of Terms Gini coefficeint is a measure of sta- Pietra index is a measure of statistical
Akaike Information Criterion is a tistical heterogenicity. heterogenicity.
method for selecting a statistical Hosmer Lemeshow (goodness-of-fit) Resampling is the statistical process of
model by estimating the relative test is a statistical method for validating prediction models by
quality of models for a given dataset. assessing the calibration of a pre- using subsets of the data. When is
Brier score assess overall performace diction model. the data is selected randomly this
of a prediction model. Link functions relate the mean of the is termed bootstrapping.
C statistic is a measure of the area response to linear predictors in the Stukel test is another method to assess
under the curve (AUC) in a model. the goodness-of-fit or calibration
receiver operating curve (ROC) Misclassification rate is a measure- of a prediction model.
and tells how well a prediction ment error of a prediction model.
model will work on average.
SUPPLEMENTAL TABLE 1
Univariate analysis to identify antenatal factors associated with CD
Characteristic Category CD rate (%) Odds ratio 95% CI P value
Maternal age, y <25 16.4 Control
2535 20.8 1.3 1.01.8 .89
>35 26.6 1.9 1.32.6 <.001
Maternal height, cm >160 18.1 Control
<160 35.8 2.5 2.03.2 <.001
Maternal BMI, kg/m2 <25 17.7 Control
2530 25.1 1.6 1.21.9 .75
>30 32.5 2.2 1.73.0 <.001
Maternal head circumference, cm <58.2 20.0 Control
>58.2 29.4 1.7 1.22.2 <.001
Ethnicity White European 20.1 Control
Other 28.1 1.5 1.02.3 .06
History of miscarriage No 21.1 Control
Yes 20.1 1.1 0.71.6 .77
Family history of CD No 19.5 Control
Yes 25.3 0.72 0.60.9 .003
Birth plan No 19.4 Control
Yes 23.4 1.3 1.01.5 .022
Education level (max level achieved) Second level 21.0 Control
Third level 21.0 1 0.81.2 .95
Alcohol use (any) in pregnancy No 20.7 Control
Yes 24.7 0.8 0.61.1 .20
Smoking status Never 22.0 Control
Ex 18.6 0.8 0.71.0 .04
Current 24.0 1.1 0.81.6 .22
Type of care Obstetric provided 21.7 Control
Midwife provided 20.3 0.9 0.81.1 .38
Fetal head circumference <90th centile 20.0 Control
>90th centile 30.0 1.7 1.32.3 <.001
Fetal abdominal circumference <90th centile 20.1 Control
>90th centile 28.8 1.6 1.22.2 .002
Estimated fetal weight <90th centile 19.9 Control
>90th centile 30.6 1.8 1.32.4 <.001
Fetal head engagement on clinical palpation (at time of Engaged 20.2 Control
study ultrasound)
Not engaged 21.4 1.1 0.91.3 .30
Uncertain 28.2 1.6 1.02.4 .07
Occiput posterior (at time of study ultrasound) No 21.5 Control
Yes 19.7 1.1 0.91.4 .37
BMI, body mass index; CD, cesarean delivery; CI, confidence interval.
Burke et al. Prediction of cesarean delivery in nulliparas. Am J Obstet Gynecol 2017.
SUPPLEMENTAL TABLE 3
Crossvalidation and bootstrap results: means and 95% confidence intervalsa
10-fold cross-validation (Repeats ¼ 100) Bootstrap resampling (repeats ¼ 1000)
Apparent Training sample Validation sample Training (bootstrap) Validation (original)
Model fit statistic fit (9/10th) (1/10th) sample sample
Downloaded for Anonymous User (n/a) at Universitas Gadjah Mada from ClinicalKey.com by Elsevier on September 14, 2017.
Cesarean delivery rate 0.21 0.21 (0.210.22) 0.21 (0.160.26) 0.21 (0.190.23) 0.21
b
Predictor correlation 0.17 0.17 (0.16, 0.18) 0.19 (0.11, 0.28) 0.17 (0.14, 0.20) 0.17
For personal use only. No other uses without permission. Copyright ©2017. Elsevier Inc. All rights reserved.
Parameter estimates
Intercept 1.46 1.46 (1.50, 1.42) N/A 1.47 (1.58, 1.35) N/A
OBSTETRICS
Maternal age (z-score) 0.19 0.19 (0.150.22) N/A 0.19 (0.080.29) N/A
Maternal BMI (z-score) 0.26 0.26 (0.220.29) N/A 0.26 (0.160.36) N/A
Maternal height (z-score) 0.54 0.54 (0.500.58) N/A 0.54 (0.420.66) N/A
AC (z-score) 0.20 0.20 (0.170.24) N/A 0.21 (0.090.32) N/A
HC (z-score) 0.24 0.24 (0.200.27) N/A 0.23 (0.130.35) N/A
Goodness-of-fit
AIC 2235 2004 (19622045) 225 (182266) 2222 (21172330) 2231 (22282239)
Hosmer-Lemeshow 10.7 8.77 (4.4014.65) 8.33 (2.4017.20) 17.75 (5.3535.75) 9.98 (4.8017.40)
c2 (8 df)
Stukel c2 (2 df) 1.2 1.34 (0.103.55) 1.58 (0.106.60) 3.36 (0.1013.00) 1.56 (0.104.35)
Overall performance
Brier score 0.14 0.15 (0.150.16) 0.15 (0.130.18) 0.15 (0.140.16) 0.15 (0.150.15)
Scaled Brier score 0.15 0.08 (0.070.09) 0.07 (0.00, 0.14) 0.08 (0.060.11) 0.08 (0.070.08)
Pietra score 0.37 0.27 (0.260.28) 0.28 (0.240.34) 0.27 (0.230.31) 0.27 (0.230.31)
Calibration
Intercept N/A N/A 0.01 (0.38, 0.32) N/A 0.00 (0.10, 0.11)
Slope N/A N/A 1.00 (0.511.59) N/A 0.98 (0.841.15)
Discrimination
D-statistic (KS) 0.34 0.29 (0.270.31) 0.33 (0.20, 0.45) 0.30 (0.250.35) 0.29 (0.280.30)
Gini coefficient 0.37 0.37 (0.360.39) 0.37 (0.210.53) 0.38 (0.320.43) 0.37 (0.360.37)
c-statistic (AUC) 0.69 0.69 (0.680.70) 0.68 (0.600.76) 0.69 (0.660.72) 0.68 (0.680.69)
Misclassification rate 0.21 0.21 (0.200.21) 0.21 (0.160.25) 0.21 (0.190.22) 0.21 (0.210.21)
AC, abdominal circumference; AIC, Akaike information criterion; AUC, area under the curve; BMI, body mass index; HC, head circumference; KS, KolmogoroveSmirnov; N/A, not available.
ajog.org
a
95% confidence intervals are the 2.5th and 97.5th percentiles of the sampling distribution. To illustrate the range of cross-validation findings, the sampling distribution includes both folds and replications from each cross-validation (10 100 ¼ 1000 samples);
b
The “predictor correlation” is the maximum multiple correlation among predictors for each sample.
Burke et al. Prediction of cesarean delivery in nulliparas. Am J Obstet Gynecol 2017.