Are Racial Disparities in Cesarean Due to Differences in Labor Induction Management?

Abstract

Objective

While there are known racial disparities in cesarean delivery (CD) rates, the exact etiologies for these disparities are multifaceted. We aimed to determine if differences in induction of labor (IOL) management contribute to these disparities.

Study Design

This retrospective cohort study evaluated all nulliparous patients with an unfavorable cervix and intact membranes who underwent IOL of a term, singleton gestation at a single institution from October 1, 2018, to September 30, 2020. IOL management was at clinician discretion. Patients were classified as Black, Indigenous, and People of Color (BIPOC) or White based on self-report. Overall rates of CD were compared for BIPOC versus White race. Chart review then evaluated various IOL management strategies as possible contributors to differences in CD by race.

Results

Of 1,261 eligible patients, 915 (72.6%) identified as BIPOC and 346 (27.4%) as White. BIPOC patients were more likely to be younger (26 years interquartile range (IQR) [22–30] vs. 32 years IQR [30–35], p < 0.001) and publicly insured (59.1 vs. 9.9%, p < 0.001). Indication for IOL and modified Bishop score also differed by race (p < 0.001; p = 0.006). There was 40% increased risk of CD for BIPOC patients, even when controlling for confounders (30.7 vs. 21.7%, p = 0.001; adjusted relative risk (aRR) 1.41, 95% confidence interval (CI) [1.06–1.86]). Despite this difference in CD, there were no identifiable differences in IOL management prior to decision for CD by race. Specifically, there were no differences in choice of cervical ripening agent, cervical dilation at or time to amniotomy, use and maximum dose of oxytocin, or dilation at CD. However, BIPOC patients were more likely to undergo CD for fetal indications and failed IOL.

Conclusion

BIPOC nulliparas are 40% more likely to undergo CD during IOL than White patients within our institution. These data suggest that the disparity is not explained by differences in IOL management prior to cesarean, indicating that biases outside of induction management may be important to target to reduce CD disparities.

There are significant, unacceptable disparities in birth outcomes between those who identify as Black, Indigenous, and People of Color (BIPOC) and those who do not in the United States.^1–3 For example, Black patients have more than a 10% increased odds of cesarean delivery (CD) compared with non-Hispanic White women, even when accounting for sociodemographic and clinical differences.^4,5 Underlying causes of health disparities such as in cesarean rate are likely multifactorial, and may relate to hospital location and variability in care quality among hospitals, different groups of practitioners providing disparate care within a given site, as well as racism and biases, both implicit and explicit.^6–9

Induction of labor (IOL), making up more than 20% of all deliveries nationally, varies widely by site and clinician across the United States.^10,11 Components of labor induction decision-making, such as frequency of cervical exams, utilization of artificial membrane rupture, oxytocin, intrauterine pressure catheters (IUPCs), and thresholds for CD, differ significantly across and within centers. This variation provides room for the impact of racial biases. Prior work has demonstrated that standardization of labor induction practices may reduce racial disparities in cesarean rate.¹²

Despite understanding that bias likely plays a significant role in cesarean disparities, we have little understanding of whether steps during the labor management process are different by race and, if, in turn, this may drive disparate obstetric outcomes. In this work, we aimed to compare labor induction management practices for nulliparous patients by race in the context of a single labor unit. This approach allowed us to control for much of the potential site or provider-related variation, to better enable us to identify whether disparities in specific steps during induction were associated with differences in CD by race, with the ultimate goal of identifying areas that could then subsequently be targeted with a disparities-focused intervention.

Methods

We performed a retrospective cohort study of all nulliparous birthing people undergoing labor induction at the Hospital of the University of Pennsylvania from October 2018 to September 2020, comparing labor induction management practices by patient race. Those patients self-identifying as Black, Asian, Native Hawaiian/Pacific Islander, Native American, Latinx ethnicity or other race were grouped as BIPOC. Patients were included if they were nulliparous and undergoing induction with a term, singleton gestation and intact membranes, requiring cervical ripening as determined by their clinical team. Patients without documented self-reported race were excluded. Patients were evaluated for inclusion/exclusion through an Electronic Health Record (EHR)-based report utilizing the above criteria, and confirmed for eligibility by either the primary investigator or one of a team of trained research assistants through individual chart review. The project was approved by the University of Pennsylvania Institutional Review Board. This manuscript was prepared utilizing the Strengthening the reporting of observational studies in epidemiology (STROBE) guidelines.¹³

During this study time frame, labor induction management was up to the discretion of the clinician without a standardized protocol in place. This time period was selected as it represented the 2-year preimplementation phase prior to a planned implementation of a standardized protocol for labor management. Our labor unit is notably staffed with a laborist model, where all presenting patients, regardless of prenatal care location, are cared for by the same team of obstetricians and trainees. CD rates were compared between BIPOC and White patients. Demographic and clinical data were collected, including indication for induction and clinician interventions during induction.

Various induction management strategies were evaluated as possible contributors to differences in CD including utilization of combined induction agents for induction initiation (defined as Foley/Cook catheter placement in addition to misoprostol placement or oxytocin initiation within 30 min of each other vs. a single agent method), cervical Foleycatheter left in place longer than recommended time frame (≥12.5 h), number of misoprostol doses utilized, cervical dilation at time of artificial rupture of membranes, time to amniotomy from induction start, use of oxytocin, maximum dose of oxytocin if utilized, time to oxytocin initiation after induction start, IUPC utilization in the setting of active labor dystocia (defined as same cervical dilation >2 h apart at ≥6 cm).

We also evaluated characteristics of deliveries performed among BIPOC and White patients, including dilation at time of cesarean if cesarean occurred, length of time in labor (latent, active, and total), and indication for and time to CD if performed. All data were collected using individual chart review by either the primary investigator or a team of trained research staff. At our institution, a failed IOL is generally defined as at least 12 hours of attempted induction while ruptured on oxytocin, or 36 hours without those requirements, without entering active labor (≥6 cm dilation), consistent with our institutional practice. Arrest of dilation and arrest of descent are defined according to ACOG’s Safe Prevention of the Primary Cesarean Delivery.¹⁴ When multiple indications for cesarean applied or were listed, the primary indication as described in the cesarean operative report was adjudicated as the primary indication.

Bivariate comparisons of demographic and predelivery clinical characteristics by patient race, as well as induction management strategies and labor outcomes were performed with Fisher’s exact tests and chi-square tests for categorical variables and t-tests or Wilcoxon rank-sum tests for continuous variables, where appropriate. Robust Poisson regression was planned to control for possible confounders of the relationship between patient race and CD. We evaluated demographic and clinical characteristics associated on bivariate tests (p < 0.20) with the exposure (BIPOC vs. White) as well as the outcome of interest (CD) as potential covariates, in addition to a priori selected variables. Using these methods, maternal age, body mass index (BMI) at the time of delivery, indication for induction, and modified Bishop score were utilized in the final model. Of note, the Wald test was utilized to determine if Bishop score should be treated as continuous or categorical; continuous was selected based on these results. All secondary analyses were performed using unadjusted bivariate analyses. Statistical analyses were performed with Stata 15 (StataCorp, College Station, TX). All tests were two-tailed, and p-values <0.05 were considered statistically significant.

Results

A total of 1,261 patients were eligible for this analysis, of which 915 (72.6%) identified as BIPOC and 346 (27.4%) as White. Demographic characteristics are shown in ►Table 1. BIPOC patients were more likely to be younger, publicly insured, and have a higher BMI. Indication for IOL and modified Bishop score also differed by race. BIPOC patients were more likely to undergo IOL for fetal indications, and less likely to undergo elective or late-term IOL. There were no significant differences in gestational age at delivery.

Table 1:

Demographic and clinical characteristics by patient race

		BIPOC (n = 915) n(%)	White (n = 346) n(%)	p-value
Maternal age (years ) a		26 [22–30]	32 [30–35]	<0.001
Insurance status	Public	538 (59.1)	34 (9.9)	<0.001
	Private	372 (40.9)	311 (90.1)
Body mass index (BMI) a		31.9 [27.5–38.1]	30.3 [27.1–34.0]	<0.001
Gestational or pregestational diabetes		92 (10.1)	34 (9.8)	0.90
Chronic hypertension		57 (6.2)	16 (4.6)	0.28
Indication for induction	Maternal b	319 (34.9)	128 (37.0)	<0.001
	Fetal c	286 (31.3)	48 (13.9)
	Elective/postdates	202 (22.2)	116 (33.5)
	Other d	108 (11.8)	54 (15.6)
Modified Bishop score a		2 [0–3]	2 [1–4]	0.006
Gestational age at delivery a		39.5 [38.6–40.4]	39.6 [38.6–40.4]	0.38

^aMedian[IQR]

^bExamples include: chronic hypertension, gestational hypertension (GHTN), preeclampsia, diabetes, renal disease, history of venous thromboembolism, cardiac disease or other chronic medical condition where induction was recommended

^cExamples include: Oligohydramnios, intrauterine growth restriction, abnormality on fetal testing

^dExamples of “other” include: history of an intrauterine fetal demise, vaginal bleeding at term, cholestasis.

BIPOC patients had a 40% increased risk of CD when compared with White patients, even when controlling for maternal age and BMI (30.7 vs. 21.7%, p = 0.001; aRR 1.41, 95% CI [1.06–1.86]; ►Table 2).

Table 2.

Logistic regression modeling to predict mode of delivery by patient race while considering potential confounders

		Incidence Rate Ratio (IRR)	Standard error	p-Value	95% Confidence interval
BIPOC race		1.41	0.20	0.016	1.06–1.86
Maternal age at delivery		1.03	0.009	<0.001	1.01–1.05
Body mass index		1.04	0.006	<0.001	1.02–1.05
Indication for induction	Maternal a	–	–	–	–
	Fetal b	1.11	0.16	0.46	0.84–1.46
	Elective/postdates	1.09	0.16	0.57	0.81–1.56
	Other c	1.01	0.19	0.94	0.71–1.45
Modified Bishop score		0.86	0.29	<0.001	0.80–0.92

Abbreviation: BIPOC, Black, Indigenous, and People of Color.

^aExamples include chronic hypertension, gestational hypertension, preeclampsia, diabetes, renal disease, history of venous thromboembolism, cardiac disease, or other chronic medical condition where induction was recommended.

^bExamples include oligohydramnios, intrauterine growth restriction, abnormality on fetal testing.

^cExamples of “other” include history of an intrauterine fetal demise, vaginal bleeding at term, cholestasis.

►Table 3 demonstrates the various induction management strategies evaluated as possible contributors to racial differences in CD. Two-thirds of patients had combination cervical ripening agents used with no difference by race. Additionally, there were no identifiable differences in IOL management by race, such as cervical dilation at or time to artificial amniotomy, use of or maximum dose of oxytocin, or dilation at CD. As shown in ►Table 4, there were also no differences in length of latent labor, length of active labor, total length of induction, or time to CD among those who underwent cesarean. However, CD was performed at an earlier dilation in BIPOC patients as compared with White patients (5 cm^4–6 vs. 6 cm,^4–10 p = 0.003). Finally, BIPOC patients were more likely to undergo cesarean for fetal indications (including nonreassuring fetal tracing) and failed IOL, while White patients were more likely to undergo cesarean for arrest of descent.

Table 3.

Secondary outcomes (induction management strategies) by self-reported race

	BIPOC (n = 915) n (%)	White (n = 346) n (%)	p-Value
Use of combined cervical ripening agent	573 (62.6)	219 (63.3)	0.83
Cervical Foley left in >12 hours	40 (4.7)	14 (4.3)	0.76
Number of misoprostol doses utilizeda	1 (1,2)	1 (1,2)	0.18
Cervical dilation at artificial amniotomya	4 (4,5)	4 (4,5)	0.05
Time from induction start to amniotomy (minutes)a	509 [337–758]	527 [350–775]	0.38
Use of oxytocin	757 (82.7)	285 (82.4)	0.88
Maximum dose of oxytocin if utilizeda	10 (6–18)	12 (6–18)	0.07
Time from induction start to oxytocin initiation (minutes)a	501 [330–723]	494 [318–692]	0.48
IUPC utilization in active labor dystociab	17 (46.0)	6 (60.0)	0.43

Abbreviations: BIPOC, Black, Indigenous, and People of Color; IUPC, intrauterine pressure catheter.

^aMedian [IQR].

^bAmong those who experienced active labor dystocia.

Table 4.

Delivery characteristics

	BIPOC (n = 915) n (%)	White (n = 346) n (%)	p-Value
Dilation at cesareana,b	5 (4–6)	6 (4–10)	0.003
Length of latent phase (minutes)a	932 [612–1,342]	895 [656–1,272]	0.90
Length of active phase (minutes)a,c	99 [11–158]	94 [0–161]	0.11
Total labor length for any mode of delivery (minutes)a	1,077 [760–1,531]	1,106 [843–1,529]	0.23
Time to cesarean (minutes)a,b	1,375 [919–1,751]	1,420 [1,016–1,829]	0.37
Indication for cesareanb
Fetal indications	133 (47.3)	24 (32.0)	<0.001
Failed induction	73 (26.0)	13 (17.3)
Arrest of dilation	38 (13.5)	10 (13.3)
Arrest of descent	17 (6.1)	23 (30.7)
Otherd	20 (7.1)	7 (9.3)

Abbreviations: BIPOC, Black, Indigenous, and People of Color.

^aMedian [IQR].

^bAmong those who experienced active labor dystocia and underwent cesarean.

^cAmong those who reached active labor.

^dOther indications for cesarean delivery included malpresentation during labor, umbilical cord prolapse, worsening hypertension remote from delivery, maternal exhaustion, and failed operative delivery.

Discussion

As seen in other work, BIPOC nulliparas were 40% more likely to undergo CD during IOL than White patients within our institution, even when controlling for other differences between groups. This disparity is not explained by differences in IOL management measured in this study. Differences were seen by race in dilation at cesarean, with CD being performed at earlier dilation in BIPOC patients, and in indication for CD, with more performed for indications that can occur earlier in the IOL process, such as fetal indications and failed IOL.

We hypothesized that by utilizing detailed chart review to evaluate for differences in induction management in a cohort where a significant racial disparity existed in nulliparous cesarean rate, we would be able to pinpoint one or more specific etiologies for the disparity. We then planned to develop targeted interventions to prevent specific differences in induction management by race, such as utilizing standardized induction protocols. However, despite a significant disparity in cesarean rate in this cohort, we were unable to identify any differences in labor induction management, prior to the decision for CD, that underlie the disparity in cesarean in this cohort.

After considering these data, especially the differences in cesarean indication without differences in induction management prior to decision for CD, these data suggest that bias in the ultimate decision for cesarean may be a target for interventions to reduce disparities in cesarean rate, rather than targeting bias in individual management steps during the IOL. Future work might determine more specifically where bias in CD decision-making occurs, such as through a blinded review of fetal heart tracings for patients who underwent cesarean for nonreassuring fetal status or a more detailed evaluation, such as through ethnography, of how decisions are made for cesarean from the perspectives of all members of the health care team. Other future directions might evaluate debiasing techniques at the time of decision for cesarean, such as moments of reflection, or individuation tools.¹⁵ Other possible solutions include shared decision-making tools for cesarean, using machine learning to help identify fetal monitoring concerns (since most of the CDs were performed for nonreassuring fetal tracing) and standardized checklists for deciding to proceed with prime cesarean, ensuring compliance with national recommendations regarding cesarean for labor dystocia or fetal distress.^16–18

This study has several strengths. It was performed only in the context of nulliparous term induction with an unfavorable cervix at a single site, limiting possible confounders in the relationship between race and cesarean. Furthermore, we were able to control for additional covariates with a possible impact on this relationship in regression modeling, including maternal age and BMI, strengthening the confidence in a disparity in cesarean rate by race. We also performed detailed chart review to evaluate for management decisions throughout many steps in the induction process that could not be captured through other data sources. These detailed data on all aspects of induction management allowed us to evaluate numerous individual induction practices, and we still did not identify specific steps that could be targeted to reduce disparities.

There are also limitations to our work. While many decisions made during the process of induction were captured in this analysis, there are other decisions we were unable to assess. For example, we did not evaluate number of cervical exams performed, amount of time spent in the room by clinicians, involvement of birth support like doulas, or attempts at fetal heart resuscitation such as patient position changes, other possible sources of biased care during IOL. While we were able to detect a significant difference in CD, we may also have been underpowered to evaluate for differences in our secondary outcomes. In addition, our site is urban and academic, with a predominantly BIPOC population, which may not be generalizable to sites with different demographics and practice models. Furthermore, given variability in induction practices by site and the laborist model utilized on our unit, it is plausible that bias could play a role in specific induction steps in other units while not seen in our own.

In conclusion, this work provides further data that there is a significant racial disparity in primary cesarean rate during labor induction. In addition, our data suggest that disparities in induction management are not likely to be driving the difference in CD. However, our data suggest that differences in decision-making about when to proceed with CD may be present, and bias outside of induction management could be targeted in future interventional studies—such as at the time of decision for cesarean.

Key Points.

• The etiologies for racial disparities in cesarean are likely multifaceted.

• In this work, there were no differences by race in measures of labor induction management.

• Biases outside of induction management during labor may be targeted to reduce CD disparities.