Hospital inclusion/exclusion criteria

Eligible community hospitals had discernible OSHPD discharge data. Hospitals were identified as community hospitals if they were not part of an integrated delivery system and did not have an obstetrician/gynecologist teaching program. Teaching status was confirmed using the Fellowship and Residency Electronic Interactive Database Access System. These hospitals were excluded because the presence of a 24-hour in-house attending physician is assumed to be the standard of care in these institutions. Hospitals were categorized as a laborist hospital if they responded that they had “≥1 physicians physically present in the hospital in specified shifts whose primary focus is to care for some or all patients in labor and delivery”, as asked in our survey ( Figure 1 ). We obtained hospital structural variables from the American Hospital Association. Hospital-level characteristics identified from the survey were evaluated as potential confounders or effect modifiers (examples under the domains and subdomains of the survey provided in Table 1 ).

Flow algorithm hospital- and patient-level inclusions for primary outcomes

Feldman. Laborists and maternal outcomes. Am J Obstet Gynecol 2015 .

Patient inclusion/exclusion criteria

OSHPD discharge data were obtained for all live births in California in 2012. Inclusion criteria included live-born, term, singleton, vertex gestations. Multiple gestations and preterm gestations were excluded. Data were linked by hospital identification numbers to the survey data and American Hospital Association report. We identified the following patient-level variables: age, race/ethnicity, and payer status. Women were categorized as laboring vs nonlaboring based on a previously published algorithm (Appendix, Supplementary Table 1 ) ( Figure 1 ). Further, based on the presence of specific codes from the International Classification of Diseases, Ninth Revision, Clinical Modification ( ICD-9-CM ) manual, women were categorized as high or low risk ( Supplementary Table 2 ). Women with a history of cesarean delivery were evaluated separately ( Figure 2 ).

Flow algorithm for hospital and patient-level inclusions for TOLAC/VBAC

TOLAC , trial of labor after cesarean; VBAC , vaginal birth after cesarean.

Feldman. Laborists and maternal outcomes. Am J Obstet Gynecol 2015 .

The total cesarean rate is stratified into a primary and repeat cesarean rate. These can be further stratified into cesarean without labor or cesarean after labor. Finally, among women with a prior cesarean, women can attempt trial of labor after cesarean (TOLAC) and have a successful vaginal birth after cesarean (VBAC) or a failed VBAC. Studies have consistently demonstrated that maternal outcomes are worse among women who attempt labor or TOLAC and fail, with increased rates of hemorrhage, hysterectomy, infection, and even death. Hence, because the term “laborist” refers to care provided on labor and delivery, the study focused on laboring women and examined the following outcomes using 2-sample t tests of hospital rates weighted by volume:

• 1.
  

  Total cesarean rate (no. of cesareans/all deliveries). We posited that fewer primary and repeat cesarean deliveries and more TOLAC attempts would lead to a lower total cesarean rate.

  
  
• 2.
  

  Primary cesarean rate (no. of cesareans/all women at risk for primary cesarean). In theory, the presence of a laborist may be associated with fewer elective or discretionary primary cesarean deliveries in addition to fewer emergent cesareans. We therefore evaluated the primary cesarean rate in the following groups:

  
  
  
  
  • a.
    

    No labor primary cesarean rate (no. of cesareans/all women at risk for primary cesarean).

    
    
  • b.
    

    Labor primary cesarean rate (no. of cesareans/all women at risk for primary cesarean who labored).

  
  
  
  
• 3.
  

  VBAC rate. There are 3 submeasures related to VBAC:

  
  
  
  
  • a.
    

    TOLAC (no. of TOLAC/all women with prior cesarean delivery): the presence of a laborist would be associated with increased rate of attempted TOLAC.

    
    
  • b.
    

    Successful VBAC (successful VBAC/all women who attempt TOLAC).

    
    
  • c.
    

    VBAC rate (successful VBAC/all women with prior cesarean delivery): the presence of a laborist should be associated with an increased VBAC rate.

  
  
  
  
• 4.
  

  Maternal composite morbidity and severe maternal morbidity rate for laboring women as described by Korst et al. These measures included conditions commonly grouped as childbirth complications ( Supplementary Tables 3 and 4 ).

All reported P values are for 2-sided tests. An adjusted P = .01 type I error rate was used due to multiple testing resulting from the consideration of multiple outcomes.

We hypothesized that, for women who labored, the presence of a laborist would be associated with a lower primary cesarean rate, a higher successful VBAC rate, and lower maternal complication rates. To evaluate the impact of the laborist staffing model on these outcomes, we developed multiple logistic regression hierarchical models using patient-level data with random intercepts for hospitals to account for intrahospital correlation in outcomes. We first fitted models using an indicator for the laborist hospitals and adjusted for patient case mix to obtain the adjusted overall (direct and indirect) effect of the presence of laborists on each of the 3 outcomes. The following patient-level variables were used for case-mix adjustments: age, race/ethnicity, insurance status, and maternal risk status (high vs low risk). Maternal risk status was based on the presence of prespecified clinical conditions. The direct effect of the presence of the laborist was then estimated by adding hospital factors that were significantly associated with the outcome to the case-mix-adjusted models. For each outcome, a forward hospital factor selection process was followed with P < .05 used for model entry. The subset of hospital factors considered as potential covariates for each outcome were those that individually tested significant (in a bivariate setting) for the outcome. Odds ratios (ORs) and 95% confidence intervals (CIs) for the effect of the presence of a laborist are reported for each regression model.

This study was approved by the California Committee for the Protection of Human Subjects (protocol 13-09-1339), and Cedars-Sinai Medical Center Institutional Review Board (protocol 00028186, expedited review).