Materials and Methods
All 1810 women who delivered live-born infants and were discharged between June 1, 1997, and January 26, 1998, from St. Joseph's Medical Center were identified through the hospital discharge data system. We excluded women with multiple gestations and women with gestations of less than 38 or at least 42 weeks. Women were considered ineligible for vaginal delivery if they had primary or secondary International Classification of Diseases, 9th Revision7 (ICD-9) codes of placenta previa or abruption (641.0–641.23), vasa previa (663.50–663.53), cord prolapse (663.0–663.03), or breech or transverse lie (652.0–652.33). Women who had induced or cesarean deliveries also were excluded if chart review revealed active herpes, history of classic or vertical incision, more than two prior cesarean deliveries, or abdominal delivery without signs of labor. We also excluded women who had scheduled primary or repeat cesarean deliveries if they had refused to deliver vaginally after a previous cesarean delivery or if no attempt was made to deliver vaginally. Women who had cesarean deliveries and were included in the study had no prelabor indications for cesarean delivery and had abdominal deliveries because of nonreassuring fetal heart rate tracings after labor had begun or because of failure to progress after a course of labor was attempted.
Women were considered ineligible for elective induction if they had primary or secondary ICD-9 codes of pregnancy-induced hypertension (642.30–642.74), insulin-dependent diabetes mellitus (648.0), gestational diabetes (648.80–648.84), fetal growth restriction (656.50, 656.51, and 656.63), oligohydramnios (658.00–658.03), polyhydramnios (657.0–657.03), chorioamnionitis (658.40–658.43), prolonged rupture of membranes (658.20–658.23), anemia (648.20–648.24), or Rh isoimmunization. Two women who had vaginal deliveries were excluded because of chart-confirmed complex medical conditions (endocarditis and severe psychosis).
Inductions were identified among all deliveries using the induction logbook, hospital discharge data (ICD-9 codes 73.01, 73.4, 96.49, and 73.1), and charge data on induction agents (dinoprostone, oral or vaginal misoprostol, and oxytocin). We reviewed the charts of all eligible women. Labors were classified as inductions only when the induction agent or method was administered or performed before there were contractions 2–5 minutes apart and before the cervix was dilated less than 3 cm. The resulting cohort comprised women with term pregnancies who delivered singleton infants in cephalic presentation, without indication for induction and without prelabor indications for cesarean delivery.
Parity information was collected from Washington State birth certificate data. Inpatient charts missing that information were reviewed and parity was recorded from prenatal records. The hospital's discharge data system provided histories of cesarean delivery (ICD-9 code 654.21), age, actual direct cost, attending provider type (certified nurse-midwife, family practitioner, obstetrician), and admission, delivery, and discharge dates and times.
Actual direct cost was determined by the hospital's decision support department using Eclipsys (Eclipsys Corp., Delray Beach, FL), a cost-accounting relative-weight method that summarizes direct costs for each patient based on clinical and ancillary resource consumption. Actual direct costs represent all nursing and supply costs generated by all contributing departments that provide service to the patient; not included are nondepartmental capital supplies, hospital administration costs, nursery-related costs, and anesthesiology fees.
All records with invalid data on hospital admission-to-delivery and delivery-to-discharge times were reviewed to determine true admission, delivery, and discharge dates and times. A quality audit of admission-to-delivery calculations was done using 73 randomly selected charts from the study period. Mothers' admission dates and times to delivery and infants' delivery dates and times were abstracted, and an admission-to-delivery time in hours was calculated for each patient. That calculation was compared with the hospital system's electronically derived admission-to-delivery time.
We used SPSS 7.5.2 (Statistical Package for Social Sciences; SPSS Inc., Chicago IL) for analysis. Categoric relationships were analyzed using the χ2 test. Categorized variables such as age and birth weight, which had a direct relationship to risk of cesarean delivery, also were tested for significance using the χ2 test for trend.8 Differences in averages were tested using the t test where appropriate (age and birth weight) and the Mann-Whitney U test for nonnormally distributed variables (parity, gestational age, cost, and inpatient predelivery and postdelivery lengths of stay).
Stepwise logistic regression model analysis was used to create a final model of risk factors for cesarean delivery, with a significance level of P < .05 for variable entry into the model. After the final model was produced, a second check of interaction terms of all variables was done, with a significance level of P < .01 for interaction terms. The Hosmer-Lemeshow statistic for goodness of fit was used to test the model.9 With that test, a model is judged to fit sufficiently well if the test assumptions are met and the P value resulting from the test is not significant (P > .05).
Of 1135 eligible subjects, 263 (23.2%) underwent elective induction of labor. One hundred seventy inductions (64.6%) were done with up to three doses of oral misoprostol 50 μg, 70 (26.6%) were done by oxytocin infusion, 20 (7.6%) by artificial rupture of membranes only, and three (1.1%) with only prostaglandin (PG) gel. Eighty-two (48.2%) of 170 misoprostol inductions also involved oxytocin infusion, and three of those also included use of PG gel. Three (4.2%) of 70 oxytocin inductions also included use of PG gel.
Women who underwent elective induction tended to be older, were more likely to have had prior cesarean deliveries, had longer gestations, were more likely to receive epidural anesthetic, were less likely to be treated by midwives, delivered larger infants, and had significantly higher cesarean delivery rates compared with women who did not undergo induction (Table 1). Although the average and median birth weights of infants of women who underwent induction did not seem clinically different from those of infants of women who did not undergo induction, the infants of women who underwent induction were statistically significantly heavier and a significantly larger proportion of these infants weighed more than 4000 g.
Initial analysis of cesarean deliveries by parity revealed an almost three-fold increased risk among nulliparas whose labor was induced compared with nulliparas whose labor was not (relative risk 2.9, 95% confidence interval [CI] 1.6, 5.2) and a two-fold increase in cesarean deliveries among parous women who underwent induction compared with parous women who did not (Table 1).
Those data showed important differences in risks of cesarean delivery based on relationships between parity and birth weight categories and parity and maternal age categories. Birth weight and maternal age appeared to have more pronounced effects on the risk of subsequent cesarean delivery for nulliparas than for primiparas or multiparas (Tables 2 and 3). Although nulliparas who underwent elective induction and delivered large infants were not at statistically significantly greater risk for cesarean than nulliparas who did not undergo induction and who delivered large infants, small numbers in the high birth weight categories might have prevented finding such an association. Cesarean deliveries were done in the case of both nulliparas who underwent induction and whose infants were heavier than 4500 g. Two of the four nulliparas who did not undergo induction and whose infants were heavier than 4500 g had cesarean deliveries. Separate regression models for nulliparas, primiparas, and multiparas were constructed based on those findings.
In the final multivariable logistic regression model predicting cesarean delivery, induction remained a significant predictor of it (OR 2.4, 95% CI 1.2, 4.9), after adjustment for the other important predictors of cesarean delivery among nulliparas only. Excluding all nulliparas with infant birth weights greater than 4000 g resulted in a model in which elective induction and gestational age were the only significant predictors of cesarean delivery (P < .01 and < .05, respectively; model not shown). The unadjusted risk estimate for cesarean delivery among nulliparas with infants who weighed more than 4000 g was the following: OR 4.2, 95% CI 2.3, 7.6 (data not shown). A revised model that included all variables in the final model (Table 4), plus the epidural anesthetic and provider type variables, did not reduce significantly the association between induction and cesarean delivery (OR 2.3, 95% CI 1.1, 4.9) (Hosmer-Lemeshow P = .80).
There were 27 cesarean deliveries among 683 parous women. For that group, the only important variable that predicted cesarean delivery was prior cesarean delivery (OR 38.1, 95% CI 14.5, 100.3). Excluding nulliparas and women who had had prior cesarean deliveries (n = 607) resulted in no variables that were significant for producing increased risk of cesarean delivery, including elective induction.
Elective inductions that ended in vaginal delivery cost $273 more (P < .001) and required an average of 4 hours more in the hospital before delivery than did noninduced vaginal deliveries (P < .001) (Table 5). The additional direct cost for elective inductions translated into $63,882 for those that ended in vaginal delivery during the 8-month study period. The additional expense and in-hospital predelivery time associated with elective inductions that ended in cesarean delivery compared with noninduced labors that ended in cesarean delivery were not statistically significant. Although the postdelivery in-hospital time was statistically significantly longer for noninduced labors that ended in vaginal delivery than for induced labors that ended in vaginal delivery, the average difference of 0.3 hours (20 minutes) is probably not important clinically or financially (Table 5).
Despite suspected widespread use of elective induction, we found only ten published studies (MEDLINE search, 1970–1998, using the terms “labor,” “labour,” “induction,” “elective,” “techniques,” and “management”) involving more than 100 subjects that addressed elective induction and a variety of outcome measures since the advent of continuous fetal monitoring and controlled oxytocin infusion.10–19 “Elective” was often defined in different ways, and consistency of findings varied according to whether nulliparas were evaluated as a discrete group. Studies that examined cesarean delivery rates without regard to parity for women who underwent elective induction yielded inconclusive results on cesarean delivery rates between subjects who had induced labor and those who did not.10–12
Studies that evaluated nulliparas as a discrete group consistently found increased cesarean delivery rates associated with elective induction,13,17–19 although these increases were not always statistically significant.13,17 In those studies, the cesarean delivery rate for nulliparas who underwent elective induction was between 6.2% and 11.7% higher than the base cesarean delivery rate for nulliparas who had spontaneous labors. In the study by Yeast et al,18 197 nulliparas who underwent elective induction had a cesarean delivery rate of 16.2%, compared with 7.9% among 4086 nulliparas who labored spontaneously. In that study, labor induction among nulliparas (indicated and elective) was the most important predictor of cesarean delivery, after adjustment for age, birth weight, and membrane status. By comparison, risk of cesarean delivery among multiparas who underwent elective induction was not significantly higher than for multiparas who labored spontaneously. Macer et al17 described a subset of 24 nulliparas with Bishop scores less than or equal to 5, who proved to be at increased risk for cesarean delivery. Although 12 of 24 subjects with low Bishop scores required cesarean deliveries, nulliparas with Bishop scores greater than 5 had a cesarean delivery rate not statistically significantly higher than that among nulliparas who did not undergo induction. In a population of nulliparas with unfavorable cervices who underwent elective induction, there was a significantly increased risk of cesarean among those who underwent induction before 42 weeks, compared with those who underwent induction at more than 42 weeks' gestation.19
Another key finding of our study was the $273 additional cost per elective induction. For our hospital, that translated into $63,882 in additional costs for women who underwent elective induction and had successful vaginal deliveries. With our MEDLINE search we were unable to locate other studies that focused on the cost-effectiveness of elective induction. Our increased costs were principally the result of increased rates of epidural anesthetic use among subjects who underwent induction and the approximate increase of 4 hours of predelivery time in the hospital. Many studies reported shorter labor times for induced deliveries compared with spontaneous ones. Our study did not measure time from induction of labor to spontaneous delivery and hence could not produce those figures.
Unmeasured or unvalidated variables related to cesarean delivery are other weaknesses of this study. Low Bishop score was one variable we were unable to record and that has been associated with increased cesarean delivery rates.17,20 Significant associations between epidural anesthetic use and cesarean delivery–related variables were reported recently in nulliparas who had spontaneous labors.21,23 Data on epidural anesthetic use were available to us, but it was impossible to tell from the charge data whether epidural anesthetic was used for labor pain or for preparation for cesarean delivery. The association between provider type and cesarean delivery or labor induction might have involved inaccurate assignments of provider type. Local data on provider type were compromised by the propensity of medical record coders to assign cesarean deliveries to obstetricians, even when a midwife had treated the labor up to the point of decision for cesarean delivery. Although we were able to adjust for epidural anesthetic use and provider type in our modeling techniques, the complete accuracy of those variables has not been assessed by chart audit. There are no reliable studies that contend that low Bishop score, epidural anesthetic use, or provider type can explain the increased risk of cesarean delivery for nulliparas who had elective induction.
Another weakness of this study was incomplete assessment of induction status within the spontaneous labor group. We reviewed most charts, but not for all of the subjects in the study. All charts for women who had cesarean deliveries or electively induced labor were reviewed, which left 835 women who had vaginal deliveries but did not undergo induction that would be noted by our detection methods. If elective inductions were missed in that group, and the misclassification was large enough, that could explain the increased risk of cesarean delivery among women who underwent elective induction. To estimate the misclassification error, we studied a random sample of 219 women who had vaginal deliveries and no induction. Applying our definition of elective induction, we found eight subjects misclassified as having had no induction, a misclassification rate of 3.65% (95% CI 0.55, 6.35). Charts for a subsample of 70 patients were reviewed by both authors to check interobserver reliability, yielding an observed agreement of 98.6% and a κ statistic of .793. If we assume a maximum error rate of 6.35%, that translates to a shift of 55 subjects from the noninduced to the induced vaginal delivery group. Our revised cesarean delivery rates would be 9.2% (induced) and 4.2% (not induced), still statistically significant (P < .01).
As it is practiced currently in our hospital system, elective induction resulted in higher direct costs per patient and increased cesarean delivery rates for nulliparas with term pregnancies who underwent elective induction, regardless of age, birth weight, provider type, or epidural anesthetic use. Thus, elective induction of labor in nulliparas should be discouraged.
1. Fleissig A. Prevalence of procedures in childbirth. BMJ 1993;306:494–5.
2. Elferink-Stinkens PM, Van Hemel OJ, Brand R. Differences in obstetrical intervention rates between Dutch hospitals. Eur J Obstet Gynecol Reprod Biol 1994;53:165–73.
3. Lampe LG. Indications for induction of labour. Eur J Obstet Gynecol Reprod Biol 1996;65:71–3.
4. Wilailak S, Saropala N, Chaturachinda K. Elective induction of labor: Ramathibodi Hospital (Jan-Jun, 1990). J Med Assoc Thai 1993;76:44–7.
5. Eggebo TM, Okland I, Braut RM. Labor induction. Quality assurance of a method [in Norwegian]. Tidsskr Nor Laegeforen 1993; 133:2686–90.
6. Wang D. Cesarean section: An analysis of 16,975 cases [in Chinese]. Chung Hua I Hsueh Tsa Chih (Taipei) 1990;70:10–2.
7. World Health Organization. International Classification of Diseases, 9th Revision. Geneva: World Health Organization, 1977.
8. Fisher LD, Van Belle G. Categorical data: Contingency tables. In: Fisher LD, Van Belle G, eds. Biostatistics: A methodology for the health sciences. New York: Wiley, 1993:246–303.
9. Kleinbaum DG, Kupper L, Muller K. Testing hypotheses in multiple regression. In: Kleinbaum DG, Kupper L, Muller K, eds. Applied logistic regression analysis and other multivariate methods. New York: Wiley, 1989:124–43.
10. Kato K, Nagata I, Furuya K, Seki K, Makimura N. Programmed induction of labor for primiparous women to ensure daytime delivery. Asia Oceania J Obstet Gynaecol 1987;13:405–15.
11. Cole RA, Howie PW, Macnaughton MC. Elective induction of labour. A randomised prospective trial. Lancet 1975;7910:767–70.
12. Vierhout ME, Out JJ, Wallenburg HC. Elective induction of labor: A prospective clinical study. I. Obstetric and neonatal effects. J Perinat Med 1985;13:155–62.
13. Prysak M, Castronova FC. Elective induction versus spontaneous labor: A case-control analysis of safety and efficacy. Obstet Gynecol 1998;92:47–52.
14. Combs CA, Singh NB, Khoury JC. Elective induction versus spontaneous labor after sonographic diagnosis of fetal macrosomia. Obstet Gynecol 1993;81:492–6.
15. Leaphart WL, Meyer MC, Capeless EL. Labor induction with a prenatal diagnosis of fetal macrosomia. J Matern Fetal Med 1997; 6:99–102.
16. Gonen O, Rosen DJ, Dolfin Z, Tepper R, Markov S, Fejgin MD. Induction of labor versus expectant management in macrosomia: A randomized study. Obstet Gynecol 1997;89:913–7.
17. Macer JA, Macer CL, Chan LS. Elective induction versus spontaneous labor: A retrospective study of complications and outcome. Am J Obstet Gynecol 1992;166:1690–6.
18. Yeast JD, Jones A, Poskin M. Induction of labor and the relationship to cesarean delivery: A review of 7001 consecutive inductions. Am J Obstet Gynecol 1999;180:628–33.
19. Wigton TR, Wolk BM. Elective and routine induction of labor. A retrospective analysis of 274 cases. J Reprod Med 1994;39:21–6.
20. Shaw KJ, Medearis AL, Horenstein J, Walla CA, Paul RH. Selective labor induction in postterm patients. Observations and outcomes. J Reprod Med 1992;37:157–61.
21. Lieberman E, Lang JM, Cohen A, D'Agostino R Jr, Datta S, Frigoletto FD Jr. Association of epidural analgesia with cesarean delivery in nulliparas. Obstet Gynecol 1996;88:993–1000.
22. Turcot L, Marcoux S, Fraser WD. Multivariate analysis of risk factors for operative delivery in nulliparous women. Canadian Early Amniotomy Study Group. Am J Obstet Gynecol 1997;176:395–402.
23. Malone FD, Geary M, Chelmow D, Stronge J, Boylan P, D'Alton ME. Prolonged labor in nulliparas: Lessons from the active management of labor. Obstet Gynecol 1996;88:211–5.