Obstetric practice in the United States currently faces a number of challenges. Cesarean delivery rates have been increasing and vaginal birth after cesarean delivery (VBAC) rates are decreasing without concomitant improvement in maternal or neonatal outcomes.1–3 Because of the inherent difficulties associated with balancing outpatient, surgical, and personal responsibilities with the unpredictable nature of labor and delivery, the laborist model has seen rapid growth in recent years and is estimated to be used in up to approximately 40% of U.S. hospitals.4,5 The term “laborist” has many definitions, but generally designates an obstetrician who provides in-house labor and delivery coverage without competing clinical duties.6 Recent studies suggest that the presence of laborists is associated with decreased rates of cesarean delivery; hypothesized reasons for these decreases include increased tolerance with equivocal fetal heart rate tracings and decreased competition between a physician's desire to patiently await a vaginal delivery and outside clinical and personal responsibilities.6–8 Access to midwifery care has also been shown to be associated with lower cesarean delivery rates, although data are sparse.9–11
We previously reported that primary cesarean delivery rates at a single institution in northern California were lower among publically insured women who delivered under a laborist model that includes 24-hour in-hospital midwifery coverage than among privately insured patients who delivered under a traditional private practice model.12 In this report we present results of an analysis of the expansion of this laborist–midwifery model to privately insured patients.
MATERIALS AND METHODS
We conducted a cohort study of all live-born deliveries at Marin General Hospital between April 2005 and March 2014. All publicly insured women at this institution, which is the only obstetric hospital in the county, are covered under the Medi-Cal program (California's state Medicaid program) and most receive prenatal care from midwives working at a public clinic, whereas the overwhelming majority of privately insured women receive their prenatal care from a private practice physician in a variety of group or solo practice settings. Before April 2011, these two groups also experienced two different models of labor and delivery care. Specifically, publicly insured women were cared for primarily by midwives with 24-hour in-house laborist back-up. Privately insured women were managed under the more conventional private practice model, in which their labor and delivery care was provided by the private physician or a covering partner who took call while at home or in the office and, in general, managed labor remotely.
In April 2011, a large change was instituted in the way that labor and delivery coverage was organized. This change occurred because of concerns about the economic feasibility of maintaining 24-hour in-house obstetrician and midwifery coverage for only half of the patients (ie, the publicly insured), the desire to expand access to midwifery services to all patients, and the difficulties faced by the private practitioners integrating labor and delivery coverage with other responsibilities.13 Coverage was changed so that the midwifery and laborist services would be available to the privately as well as the publicly insured women. A group of 10 private practice physicians began to take labor and delivery call as laborists, joining the obstetricians who had been caring for the publicly insured (those providing prenatal care to the publicly insured women as well as per-diem laborists). After the change, one midwife and one laborist were in house, 24 hours a day, working collaboratively to provide primary labor management for all private and public patients of the newly formed group. For the private patients, this new model introduced two significant changes: the option of midwifery care and having their managing clinician in house at all times. Although privately insured women who opted for physician care could request to be delivered by a member of the private practice group if the in-house obstetrician was not a member, their labor was always managed by the in-house laborist. A small group of private physicians did not participate in this change and maintained the private practice model; the patients cared for by these health care providers were not included in this analysis. Our primary hypotheses were that this expansion would be associated with 1) decreased primary cesarean delivery rates and 2) increased VBAC rates among the privately insured patients.
Maternal and neonatal data were collected prospectively by delivering clinicians and nurses as well as three dedicated chart abstractors. These data were used for state reporting and birth certificate reporting as well as research, ensuring complete records.
The primary predictor for this analysis was the change in practice occurring in 2011 that expanded access to 24-hour in-house midwifery and laborist services to the privately insured women whose health care providers participated in the laborist system. Our two primary outcomes were primary cesarean delivery rates among nulliparous women carrying term, singleton pregnancies in the vertex position and VBAC rates among women with a history of prior cesarean delivery. To examine the possibility of secular trends in primary cesarean delivery and VBAC rates over time, we also examined these outcomes in the publicly insured women whose care model did not change and who thus served as a nonequivalent control group.14,15 As a secondary outcome, we examined a composite of short-term adverse neonatal outcomes including 5-minute Apgar score less than 7, umbilical artery pH less than 7.0, and umbilical artery base excess less than −12.
Demographic and clinical characteristics were compared before and after the practice change using χ2 or t tests; this was done separately for the privately insured and publicly insured patient groups.
We analyzed changes in the outcomes three ways: 1) a simple pre–post analysis comparing rates before and after the practice change, stratified by insurance status and using χ2 tests; 2) a multivariable logistic regression model comparing rates before and after the change, controlling for maternal clinical and demographic covariates and including an interaction term with insurance status; and 3) an interrupted time-series analysis that allowed us to examine the possibility of concurrent secular trends in cesarean delivery rates, which could influence the results. Interrupted time-series models estimate the slope of each outcome over time before and after the intervention, compare the slope of outcome rates before and after the practice change, and examine the model-predicted change in outcome level at the time of the expansion. For each outcome, interrupted time-series models were fit as linear models of annual rates expressed as percentages with an autoregressive residual correlation structure. Statistical significance was considered with P values <.05; interactions with P values <.1 were examined more closely. All analyses were performed with SAS 9.3.
Deliveries included in these analyses occurred between April 2005 and March 2014 (6 years before and 3 years after the expansion) and were among women who were either covered by public insurance and received prenatal care at the publicly funded clinic or who had private insurance and were cared for by the private practice physicians who joined the laborist system. Approval for the study was granted by the University of California, San Francisco Committee on Human Research and the Marin General Hospital institutional review board.
Between April 2005 and March 2014, there were a total of 13,194 births for a mean annual delivery volume of 1,466. Of these, 4,884 deliveries met criteria for inclusion in the analysis, 3,413 before the practice change and 1,471 afterward. Three thousand five hundred sixty women were nulliparous and delivered term, singletons in the vertex position, and 1,324 had a prior cesarean delivery. Approximately half (49%) of these women (n=2,406) were privately insured and, as a group, were exposed to the practice change. There were 10 private obstetricians who changed their practice model and began providing 24-hour in-house laborist coverage without any change in reimbursement scheme, because they were salaried and were paid the same for a shift whether or not they performed any deliveries. Before the change, laborist coverage was provided by three obstetricians who provided prenatal care to the publicly insured women covering 25% of the shifts, a laborist program director who covered 15% of the shifts, and 10–12 per-diem laborists who covered the remaining 60% of the shifts. After the change, the 10 private practice physicians who joined the laborist call pool covered 30% of the shifts, and the number of shifts covered by six to eight per-diem laborists went down to 30%. Twenty-four–hour in-house midwifery coverage was provided by an annual average of 14 midwives before and 12 midwives after the expansion with the decrease in staff resulting from an increase in full-time rather than part-time workers. The midwifery service attended to 21% of the privately insured vaginal births in the first year of the new program and increased to 42% by the third year. In contrast, the midwifery service attended 80–90% of the vaginal births among the publicly insured before and after the change.
There were several differences in the demographic and clinical characteristics of the privately insured women who delivered after the practice change compared with those who delivered before, including a higher percentage with diabetes and hypertension and a lower percentage who underwent induction of labor (Table 1). Among publicly insured women, we observed increases in the percentage who had college degrees, the percentage who used epidurals, and maternal age.
The privately insured nulliparous women whose health care providers participated in expansion of midwifery and laborist services had a decrease in the rate of primary cesarean delivery from 31.7% to 25.0%, (P=.005; Table 2). During the same time period, the primary cesarean delivery rate did not significantly change among the publicly insured women (15.5% compared with 16.1%, P=.78).
A multivariable logistic regression model (controlling for maternal age, race–ethnicity, education, gestational age, epidural, induction of labor, maternal medical complication, birth weight, delivery year, and with an interaction term for insurance type) revealed a clinically and statistically significant reduction in primary cesarean delivery rates after the practice change among the privately insured women (adjusted odds ratio [OR] 0.56, 95% confidence interval [CI] 0.39–0.81, P=.002) but not among the publicly insured women (adjusted OR 0.80, 95% CI 0.54–1.17, P=.25). The P value for the interaction term was .092.
Similar to national trends, the primary cesarean delivery rate among the privately insured increased by 0.6% per year before the practice change (Fig. 1). During the first year of the expansion the primary cesarean delivery rate dropped by an estimated 6.9% (P=.009) and thereafter this rate continued to decrease by nearly 2% per year. The slopes of the estimated annual change in primary cesarean delivery rates both before and after the practice change significantly differed from zero, and the two slopes significantly differed from one another (Appendix 1, available online at http://links.lww.com/AOG/A681).
The primary cesarean delivery rates among the publicly insured women followed a similar trend, but although the regression slope before the expansion was significantly different from zero, the slope after the expansion was not, and the rates of change before and after were not significantly different from each other (Appendix 1, http://links.lww.com/AOG/A681). There was no significant drop at the time of expansion among the publicly insured. We performed a sensitivity analysis including induction of labor as a covariate because the rates of induction changed during the study period and found results similar to the primary analysis among the privately insured women: the decrease at the time of the practice change was estimated at 3% (P=.029) and the annual primary cesarean delivery rate was 3% lower after the expansion than before (P=.004).
We also observed a substantial increase in the VBAC rate among the privately insured women who had a prior cesarean delivery, from 13.3% to 22.4% (P=.002; Table 3). Whereas the VBAC rate decreased among the publicly insured women, this change was not statistically significant (33.8% compared with 26.8%, P=.07). When we adjusted for confounders, the increase in VBAC rate remained statistically significant in the privately insured group (adjusted OR 2.03, 95% CI 1.08–3.80, P=.03). The P values for the interaction between the pre–post practice change indicator and insurance status were <.001 and <.002 for the unadjusted and adjusted models, respectively.
Results of the interrupted time series analysis of trends in VBAC rate over time before and after the practice change are shown in Figure 2. Before the change, there was no significant effect of calendar year on the VBAC rate among the privately insured women (0.4% annual decrease, P=.38). After the practice change, privately insured women had an 8% annual increase in VBAC rate (P=.004); the difference between pre–practice change and post–practice change slopes was statistically significant (P=.004) (Appendix 2, available online at http://links.lww.com/AOG/A681). Among the publically insured women, the difference in model-predicted VBAC rates in the year after the practice change was not significant, and no difference emerged in the decreasing rate of VBAC before and after the practice change (−2% compared with −4%, P=.3 for difference pre–practice change compared with post–practice change slopes). In other words, although the simple pre–post analysis of publically insured women suggested a lower rate of VBAC after the expansion, the interrupted time-series analysis revealed that this change in rate represented a persistent trend rather than a change that might be attributable to the intervention itself.
There were no statistically significant differences among rates of the composite of adverse short-term neonatal outcomes (incidence of Apgar score less than 7, umbilical artery pH less than 7.0, or umbilical artery base excess greater than 12) before or after the expansion among the privately insured women exposed to the practice change (21 of 1,649 [1.3%] compared with 14 of 748 [2.3%], P=.07) (Table 4). As seen in Figure 3, adverse neonatal outcomes among the privately insured patients were increasing before the practice change and maintained a similar rate of increase afterward without any statistically significant change after the expansion (0.07% compared with 0.66%, P=.11 for differences in rates of change) (Appendix 3, available online at http://links.lww.com/AOG/A681). The interaction between the pre–post practice change indicator and insurance status was not statistically significant in either the adjusted or unadjusted model (P=.14 for both models).
In this study, we found that the expansion of midwifery and laborist services in a collaborative practice model at a single community hospital was associated with a decreased primary cesarean delivery rate and an increased rate of VBAC.
Our study has two methodologic strengths that bolster the validity of our findings. First, unlike many pre–post analyses of practice changes at the hospital level, we had a control group at the same hospital that would have experienced any other hospital-wide interventions that may have had an effect on cesarean delivery rates. Also, we performed an interrupted time-series analysis in addition to the simple pre–post analysis. Although a pre–post analysis is easier to understand, it can lead to misleading results. For example, if the primary cesarean delivery rates were steadily decreasing in this hospital as a result of other factors, a simple pre–post analysis would estimate a lower rate after the expansion, but this would not be attributable to the practice change. In our study, we graphically demonstrated the converse: primary cesarean delivery rates were increasing slightly before the expansion and decreased afterward. This pattern of change in the primary cesarean delivery rates was similar in both groups, although the magnitude of the change was only clinically and statistically significant among the privately insured women.
Another important finding is the substantial increase in VBAC rates among the privately insured women. Only the group of women whose health care providers participated in this expansion saw a reversal of the trend of decreasing VBAC rates with an increase in VBAC rate, both at the time of the expansion and for each year thereafter. The presence of 24-hour in-house health care providers who were directly involved in the care of privately insured women may explain this increase; private prenatal care providers may have been more likely to recommend a trial of labor after cesarean delivery because they no longer had to contend with the difficulties of managing these high-risk patients remotely or being called in to attend to an emergency while on call from the office or home.
A similar decrease in primary cesarean delivery rates associated with implementation of a laborist program in Nevada has been reported; however, the Nevada study only demonstrated decreased rates with dedicated laborists without any outpatient duties.7 The program described in our analysis differed from the Nevada program in two ways: midwifery services were also included and the in-house laborists cared for all laboring patients regardless of insurance status, which may explain why we observed a decrease in cesarean delivery rates using “community laborists” while they did not.
Although our results point to a clear decrease in primary cesarean delivery and increase in VBAC rates associated with the midwifery–laborist expansion, we are unable to identify whether this change was the result of the increased availability of in-house obstetricians and midwives, the collaborative practice model that developed as a result of the expansion, or another unidentified factor. We cannot be certain whether other factors could have led to the decrease in rates, although there were no other official hospital policies that took effect during this time. It is possible that the decreased induction rate was the result of fewer elective inductions for health care provider convenience, but the sensitivity analysis that included induction status suggests that any change in induction rates was not the primary driver of the decrease in cesarean delivery seen in our study.
Several limitations deserve comment. This is an analysis of a practice change at a single institution; whether similar changes would happen with the implementation of this model at other hospitals is not known. As a result of the rarity of adverse events and a high degree of missing cord gas collection (40%), we were underpowered to detect small changes in the rates of adverse neonatal outcomes. Based on our sample size and the low frequency of adverse events, our study was powered to detect a postintervention rate in the privately insured of 3.0%, and the 95% CI calculated from the interrupted time series suggests that the true difference in slopes before and after the intervention may range from −0.02% to 1.20%. We therefore cannot rule out a true small increase in adverse neonatal outcomes, and future studies should focus efforts on complete data collection of adverse neonatal outcomes including neonatal intensive care unit admission, birth trauma, and routine cord gas assessment.
Our study adds to the growing literature on the association between laborists and midwives and decreasing primary cesarean delivery rates and increasing VBAC rates. As other maternity hospitals consider changing their staffing paradigms, similar analyses should be conducted to evaluate the association of expanding laborist and midwifery services with maternal and neonatal outcomes.
1. Podulka J, Stranges E, Steiner C. Hospitalizations related to childbirth, 2008. HCUP statistical brief #110. Rockville (MD): Agency for Healthcare Research and Quality; 2011.
2. March of Dimes peristats. March of Dimes; 2013. Available at: www.marchofdimes.com/peristats
. Retrieved May 31, 2013.
3. Callaghan WM, Creanga AA, Kuklina EV. Severe maternal morbidity among delivery and postpartum hospitalizations in the United States. Obstet Gynecol 2012;120:1029–36.
4. Srinivas SK, Shocksnider J, Caldwell D, Lorch S. Laborist model of care: who is using it? J Matern Fetal Neonatal Med 2012;25:257–60.
5. Weinstein L. The laborist: a new focus of practice for the obstetrician. Am J Obstet Gynecol 2003;188:310–2.
6. Committee Opinion no. 459: the obstetric-gynecologic hospitalist. Obstet Gynecol 2010;116:237–9.
7. Iriye BK, Huang WH, Condon J, Hancock L, Hancock JK, Ghamsary M, et al.. Implementation of a laborist program and evaluation of the effect upon cesarean delivery. Am J Obstet Gynecol 2013;209:251.e1–6.
8. Srinivas SK, Lorch SA. The laborist model of obstetric care: we need more evidence. Am J Obstet Gynecol 2012;207:30–5.
9. Johantgen M, Fountain L, Zangaro G, Newhouse R, Stanik-Hutt J, White K. Comparison of labor and delivery care provided by certified nurse-midwives and physicians: a systematic review, 1990 to 2008. Womens Health Issues 2012;22:e73–81.
10. Davis LG, Riedmann GL, Sapiro M, Minogue JP, Kazer RR. Cesarean section rates in low-risk private patients managed by certified nurse-midwives and obstetricians. J Nurse Midwifery 1994;39:91–7.
11. Harvey S, Jarrell J, Brant R, Stainton C, Rach D. A randomized, controlled trial of nurse-midwifery care. Birth 1996;23:128–35.
12. Nijagal MA, Kuppermann M, Nakagawa S, Cheng Y. Two practice models in one labor and delivery unit: association with cesarean delivery rates. Am J Obstet Gynecol 2015;212:491.e1–8.
13. Nijagal MA, Wice M. Expanding access to midwifery care: using one practice's success to create community change. J Midwifery Womens Health 2012;57:376–80.
14. Cook TD, Campbell DT. Quasi-experimentation: design and analysis issues for field settings. Boston (MA): Houghton Mifflin Company; 1979.
15. Shadish WR, Cook TD, Campbell DT. Experimental and quasi-experimental designs for generalized causal inference. Boston (MA): Houghton Mifflin; 2002.