The incidence of induction has increased significantly over the past decade, prompting increased attention and criticism, especially as it relates to elective delivery for patient or physician convenience. Initially, concerns were raised over iatrogenic prematurity with inductions or cesareans <39 weeks. Our large multicenter cohort of medical records confirms other smaller, single-center studies showing that some neonatal outcomes improve until 39 weeks regardless of labor onset type. Our data suggest that neonatal outcomes are gestational age dependent regardless of labor onset type.
It is clear that an elective induction or unlabored cesarean not meeting ACOG gestational age criteria of at least 39 weeks of gestation is at increased risk of suboptimal neonatal outcome. However, given the increased utilization of elective induction, it is encouraging that within a given gestational age, elective induction of labor does not worsen neonatal outcomes. In fact, we found that infant ventilator use, sepsis, and NICU admissions were less likely with elective induction than spontaneous labor for a given gestational age.
Concerns over maternal outcomes in elective induction of labor have previously focused on length of labor and the increased risk of cesarean delivery. We chose to look at additional maternal morbidity that we believed represented more profound maternal morbidity. Elective induction was associated with a lower risk of maternal ICU admission, and unlabored cesarean was associated with an increased risk. The decreased risk of maternal ICU admission associated with elective induction is likely due to the fact that anyone with underlying comorbidity would likely be considered an indicated induction (not an elective induction). Elective inductions, by definition, occur among healthy women to start. Indicated induction, but not elective induction, was associated with an increased odds ratio for endometritis. As the induction process is similar in both populations, it is likely that some of the indicated inductions were for chorioamnionitis. We also found that unlabored cesarean overall and elective induction at term were associated with increased risk of hysterectomy.
Our finding that hysterectomies are significantly increased in unlabored cesarean overall and term elective induction of labor is important in counseling patients about elective delivery. While much morbidity is recoverable and does not have a lasting impact on a woman’s life, losing her uterus may have a profound impact on her family plans and may be weighed differently than a cesarean delivery, or a postpartum infection. The advantages of an elective delivery are the convenience of being able to plan delivery and perhaps more control over who is the delivering provider. These advantages pale in comparison to 3.21 times the risk of hysterectomy at term for an elective induction or 6.57 increased risk for unlabored cesarean at term. We recognize, however, that the association with increased hysterectomy risk is based on only 24 hysterectomies in our final dataset and that absolute rates of hysterectomy remain low. This highlights the need for large multicenter datasets such as ours to examine peripartum hysterectomies and other rare childbirth outcomes.
Our study population was designed to represent a low-risk obstetric population to represent the average-risk patient considering elective induction/cesarean delivery. Hence, we limited our dataset to vertex, singletons, without a prior uterine scar and without conditions that clearly increase risk of hemorrhage (previa and abruption). Furthermore, we tried to account for common morbidity in our model. While the percent of our population that is completely “elective” cesarean is unknown, we defined our unlabored cesarean population as those who had no induction, ≤ 2 vaginal examinations, and a cesarean for their delivery mode. Thus, it is possible we included some women in the unlabored cesarean group who in fact had cesareans after the onset of spontaneous labor or presented to labor and delivery with an immediate need for delivery such as nonreassuring fetal status. Some may have other underlying issues precluding labor predisposing them to hysterectomy, for example large obstructing fibroids. However, our multicenter dataset with predetermined data fields that contain clinical information such as reason for admission, examination on admission, number of examinations, type of medications, as well as specified maternal and neonatal outcomes is an improvement from previous single-center studies based on chart review, or multicenter studies where delivery method and patient outcomes are based on administrative data.
Although we did not report postpartum hemorrhage rates, or rates of transfusions, our findings add to the evidence that elective induction of labor increases hemorrhage risk as evidenced by increased hysterectomies. This is consistent with a recent publication from the Norway birth registry showing that hemorrhage from atony is increased with elective induction of labor.6
The reasons for the observed increase in atony are not clear. We could speculate that inductions may have longer labors, and increase use of uterotonic agents all of which may predispose to postpartum hemorrhage.
Using EMRs for clinical research enhances efficiency in data collecting, but combining EMRs from different institutions can be difficult if data are not entered in a consistent way across sites. For example, if a clinician puts free text in a note stating the patient had a postpartum fever and endometritis, this may not be recognized as endometritis by the system if the discrete data fields “postpartum fever” or “endometritis” are not checked off or entered correctly. Similarly, our construction of the elective induction variable depends on the appropriate charting in the medical record. If a physician is inducing labor for a reason but fails to document that reason, it would be considered an unknown reason for induction and dropped from our study. It is difficult to estimate how often physicians are not documenting appropriately in the medical record in general and the EMRs in particular, as the medical record is generally considered to be the gold standard of what has happened. However, from clinical experience, it is clear that physicians do “underdocument” or do not explicitly specify all reasons for induction. Our data suggest that this may be happening since we show a substantial number of elective deliveries <37 weeks–a time when early deliveries are more likely to be indicated. Thus, we suspect our elective induction variable likely contains some women who truly had indicated inductions of labor. This mixing of indicated and elective deliveries likely mutes the true differences between elective induction and indicated induction.