Study design

We performed a cohort study to compare pregnancy outcomes of women delivering at unexposed (nonlaborist) hospitals vs those delivering at exposed (laborist) hospitals, using data from NPIC/QAS from 1998 through 2011. The NPIC/QAS is a voluntary benchmarking organization that began in 1985 with a charter membership of major perinatal centers across the United States. Within the group, the average annual delivery volume was 4619 per hospital with a range of 589-16,544 annual deliveries per hospital. The characteristics of women delivering at member hospitals represent the general US population with 70% between the ages of 21-35 years and approximately 40% unmarried. At the time the 2010 cross-sectional survey was performed, NPIC/QAS had 74 member hospitals from 26 states. NPIC/QAS has key contacts at each of their member institutions. The survey was completed by those identified at each institution to be the best qualified to complete the instrument.

Exposure status (implementation of laborists) during the time frame and the timing of exposure implementation was based on the response to a specific question on their cross-sectional survey: “Do hospitalists/laborists perform deliveries?” Sixteen NPIC/QAS member hospitals indicated on their member survey that hospitalists/laborists were performing deliveries in 2011. NPIC/QAS staff approached these hospitals to participate in the study, which verified the accuracy of the exposure (laborist/nonlaborist) designation and determined when the hospitals initiated the laborist system at their institution. Of these 16, 8 hospitals made up the final cohort. Four hospitals expressed interest but were unable to obtain internal buy-in prior to closure of enrollment, and 1 laborist hospital declined. The 3 remaining hospitals agreed to participate, but they did not have sufficient data after the implementation of the laborist program for inclusion in the study. The 8 laborist hospitals in this study implemented laborists in the years 2000, 2004, 2006 (2), 2007, 2008, 2009, and 2010.

We then used a matched sampling methodology to select eligible hospitals from the NPIC/QAS membership. Matched sampling is a cost-efficient way of comparing a treatment (laborist) group to a control (nonlaborist) group when there are considerably more control units than treated units; by matching multiple control units to each treated unit, we obtain almost as powerful a study as if we had used all control units at a fraction of the cost. Hospitals were matched 2:1 nonlaborist to laborist using the variables:

• 1)
  

  Annual volume of deliveries categorized as ≤1000 or >1000.

  
  
• 2)
  

  Geography based on US census bureau designated areas: Northeast, Midwest, South, West.

  
  
• 3)
  

  Teaching hospital status (presence of obstetrics residents).

  
  
• 4)
  

  Level of neonatal intensive unit care.

Each of these factors has been associated with changes in maternal/neonatal outcomes, the possibility of moving to a laborist system, or both. Thus, including these criteria in our match minimized other differences between laborist and nonlaborist hospitals that may be associated with maternal or neonatal outcomes.

Of the 16 nonlaborist NPIC/QAS member hospitals approached as potential matches for the laborist sites, only 1 hospital declined to participate and was replaced with the second choice hospital. After hospitals consented, maternal and neonatal discharge data from all 24 participating hospitals were obtained in a de-identified data set. The 3 hospitals in each triad (1 laborist and 2 nonlaborists) each contributed at least 3 years of data in the preimplementation period and up to 2 years of data in the postimplementation period, with the assignment of the periods for each triad based on the calendar year of laborist hospital implementation (implementation is year 0) within that triad. Available NPIC/QAS data included patient-level data submitted on all perinatal discharges by member hospitals for each quarter. NPIC/QAS processed the data and each hospital signed off on data accuracy prior to its final inclusion in the NPIC/QAS data set. The file was composed of discharge abstract/UB 04 data. Per NPIC/QAS protocol, maternal hospitalizations were linked using either medical record or billing number to the corresponding infant hospitalization. The primary investigator and analysis team were blinded to hospital identity.

Outcomes and covariates

Outcome measures were chosen based on their public health relevance, measures of patient safety, or measures of the relative health of the mother or infant. The International Classification of Diseases, Ninth Revision ( ICD-9 )- Clinical Modification ( CM ) codes or other data fields used to identify each outcome are listed in Table 1 . Maternal outcome measures included pregnancy complications (eg, postpartum hemorrhage [defined by ICD-9 codes corresponding to an estimated blood loss >500 mL for a vaginal delivery or >1000 mL for a cesarean delivery], infection, or need for intensive care admission); Agency for Health Care Research and Quality patient safety indicators such as significant perineal lacerations; and preterm birth, both spontaneous and medically indicated. Preterm birth was defined using diagnosis codes for preterm birth <37 weeks. Medically indicated preterm births were defined as those with the diagnoses codes for intrauterine growth restriction, pregnancy-related hypertension, and previa/abruption. Neonatal outcomes included birthweight at delivery, mortality, birth injury, and neonatal intensive care admission.

Final outcomes were chosen after 1 additional step. To minimize the bias of not attributing an already existing trend to the implementation of laborists, we examined the difference in each outcome measure between laborist and their matched nonlaborist hospitals during the 3 years before the implementation of the laborist program. We excluded any outcome measures where there was a statistically significant change between laborist and nonlaborist hospitals during the 3-year preimplementation time period. Including these measures could inappropriately attribute the change to the initiation of the laborist model at that hospital, when in fact there was an already existing secular trend prior to laborist implementation. The final maternal and neonatal outcomes evaluated include: cesarean delivery, chorioamnionitis, induction of labor, preterm birth, maternal prolonged length of stay (>2 days postpartum for vaginal delivery; >4 days postpartum for cesarean delivery), Apgar at 5 minutes of <7, birth asphyxia, injury, trauma, and neonatal death.

Measured covariates used for risk adjustment were those demonstrated to be associated with the exposure or outcome. These included socioeconomic variables such as insurance status; year of delivery; maternal comorbid conditions and complications around delivery such as hypertension and diabetes; and birthweight for neonatal outcomes. ICD-9-CM codes and other specified data fields were used to identify these covariates ( Table 1 ).

Statistical approach

We used a before-and-after study with an untreated comparison group to examine whether the implementation of the laborist model at a hospital was associated with a change in the underlying trend in patient outcomes in the hospital, also known as a difference-in-differences approach. Here, we examined how the baseline risk-adjusted rate of each outcome changed in the laborist hospital after implementation of this program, compared to the change in such outcomes at the same time in similar hospitals that did not implement such a program. The before-and-after study design with an untreated comparison group analyzed with the method of difference-in-differences improves the ability to determine causality by preventing bias from 3 possible sources. First, a difference between laborist and nonlaborist hospitals that is stable over time cannot be mistaken for an effect of the introduction of the laborist, because data from both the preimplementation and postimplementation period are included in the model. Second, by including year indicators in the logistic model, changes over time that effect all hospitals similarly such as overall national increase in operative delivery cannot be mistaken for an effect of the laborist. Third, the effect of differential changes to the mix of patients at different hospitals will not be attributed to the implementation of the laborist model if these changes are accurately reflected in measured risk factors.

For each outcome of interest, we fitted a logistic regression with an interaction term of laborist status and an indicator for before/after the change, after adjusting for fixed effects for hospital, year, and patient risk factors. Huber-White robust SE (synonymous with generalized estimating equation sandwich SE) accounted for clustering by hospital (the nonindependence of patients treated at the same hospital). The effect of the laborist, presented as odds ratios (ORs), measured the degree to which the outcome changed in laborist vs nonlaborist hospitals after adjusting for confounders.

Study design

We performed a cohort study to compare pregnancy outcomes of women delivering at unexposed (nonlaborist) hospitals vs those delivering at exposed (laborist) hospitals, using data from NPIC/QAS from 1998 through 2011. The NPIC/QAS is a voluntary benchmarking organization that began in 1985 with a charter membership of major perinatal centers across the United States. Within the group, the average annual delivery volume was 4619 per hospital with a range of 589-16,544 annual deliveries per hospital. The characteristics of women delivering at member hospitals represent the general US population with 70% between the ages of 21-35 years and approximately 40% unmarried. At the time the 2010 cross-sectional survey was performed, NPIC/QAS had 74 member hospitals from 26 states. NPIC/QAS has key contacts at each of their member institutions. The survey was completed by those identified at each institution to be the best qualified to complete the instrument.

Exposure status (implementation of laborists) during the time frame and the timing of exposure implementation was based on the response to a specific question on their cross-sectional survey: “Do hospitalists/laborists perform deliveries?” Sixteen NPIC/QAS member hospitals indicated on their member survey that hospitalists/laborists were performing deliveries in 2011. NPIC/QAS staff approached these hospitals to participate in the study, which verified the accuracy of the exposure (laborist/nonlaborist) designation and determined when the hospitals initiated the laborist system at their institution. Of these 16, 8 hospitals made up the final cohort. Four hospitals expressed interest but were unable to obtain internal buy-in prior to closure of enrollment, and 1 laborist hospital declined. The 3 remaining hospitals agreed to participate, but they did not have sufficient data after the implementation of the laborist program for inclusion in the study. The 8 laborist hospitals in this study implemented laborists in the years 2000, 2004, 2006 (2), 2007, 2008, 2009, and 2010.

We then used a matched sampling methodology to select eligible hospitals from the NPIC/QAS membership. Matched sampling is a cost-efficient way of comparing a treatment (laborist) group to a control (nonlaborist) group when there are considerably more control units than treated units; by matching multiple control units to each treated unit, we obtain almost as powerful a study as if we had used all control units at a fraction of the cost. Hospitals were matched 2:1 nonlaborist to laborist using the variables:

• 1)
  

  Annual volume of deliveries categorized as ≤1000 or >1000.

  
  
• 2)
  

  Geography based on US census bureau designated areas: Northeast, Midwest, South, West.

  
  
• 3)
  

  Teaching hospital status (presence of obstetrics residents).

  
  
• 4)
  

  Level of neonatal intensive unit care.

Each of these factors has been associated with changes in maternal/neonatal outcomes, the possibility of moving to a laborist system, or both. Thus, including these criteria in our match minimized other differences between laborist and nonlaborist hospitals that may be associated with maternal or neonatal outcomes.

Of the 16 nonlaborist NPIC/QAS member hospitals approached as potential matches for the laborist sites, only 1 hospital declined to participate and was replaced with the second choice hospital. After hospitals consented, maternal and neonatal discharge data from all 24 participating hospitals were obtained in a de-identified data set. The 3 hospitals in each triad (1 laborist and 2 nonlaborists) each contributed at least 3 years of data in the preimplementation period and up to 2 years of data in the postimplementation period, with the assignment of the periods for each triad based on the calendar year of laborist hospital implementation (implementation is year 0) within that triad. Available NPIC/QAS data included patient-level data submitted on all perinatal discharges by member hospitals for each quarter. NPIC/QAS processed the data and each hospital signed off on data accuracy prior to its final inclusion in the NPIC/QAS data set. The file was composed of discharge abstract/UB 04 data. Per NPIC/QAS protocol, maternal hospitalizations were linked using either medical record or billing number to the corresponding infant hospitalization. The primary investigator and analysis team were blinded to hospital identity.

Outcomes and covariates

Outcome measures were chosen based on their public health relevance, measures of patient safety, or measures of the relative health of the mother or infant. The International Classification of Diseases, Ninth Revision ( ICD-9 )- Clinical Modification ( CM ) codes or other data fields used to identify each outcome are listed in Table 1 . Maternal outcome measures included pregnancy complications (eg, postpartum hemorrhage [defined by ICD-9 codes corresponding to an estimated blood loss >500 mL for a vaginal delivery or >1000 mL for a cesarean delivery], infection, or need for intensive care admission); Agency for Health Care Research and Quality patient safety indicators such as significant perineal lacerations; and preterm birth, both spontaneous and medically indicated. Preterm birth was defined using diagnosis codes for preterm birth <37 weeks. Medically indicated preterm births were defined as those with the diagnoses codes for intrauterine growth restriction, pregnancy-related hypertension, and previa/abruption. Neonatal outcomes included birthweight at delivery, mortality, birth injury, and neonatal intensive care admission.

Table 1

Codes for outcomes and comorbid conditions

Variable name	Identifying ICD-9 codes if applicable
Maternal outcomes
Induction of labor rate	73.01, 73.1, 73.4
Cesarean delivery rate	654.2, 669.7, 370, 371
Complications of labor induction	763.7, 763.82
Transfer to ICU	Calculated
Prolonged length of stay	Calculated
Postpartum hemorrhage/blood transfusion	666.0, 666.1, 666.2, 99.0X
Chorioamnionitis/endometritis	658.4, 670, 646.6
Wound infection	674.1 (Cesarean), 674.2 (vaginal)
Third- or fourth-degree perineal lacerations	664.2, 664.3
Modified Adverse Outcome Index	Calculated
ICU admission	Calculated
Days in hospital	Calculated
Hospital readmission within 1 wk of discharge	Calculated
Neonatal outcomes
NICU/intermediate care admission	Submitted
1- and 5-min Apgar scores	Submitted
Neonatal mortality	Calculated from discharge disposition
Fetal mortality	779.9, 798.1, 669.9
Preterm delivery	644.2
Birth injury (nerve and other)	767.5, 6, 7, 8, 9
Birth asphyxia	768.1, 3, 4, 5, 6, 7, 768.9
Prolonged length of stay	Calculated
Necrotizing enterocolitis	777.5
Bacterial sepsis	771.81, 771.83
Meningitis	Calculated from multiple types
Any fracture	767.2 (Clavicle), 767.3 (other), 767.4 (spine)
Comorbid conditions
Birthweight	Numeric or fifth digit 764 or 765
Gestational age	765.2 or Submitted
Maternal marital status	Submitted
Maternal race	Submitted
Maternal insurance status	Submitted
Gravida	Submitted
Parity	Submitted
Disorders of placentation	641.2, 641.0, 762.0, 762.1
Pregnancy-induced hypertension	642.3
Chronic hypertension	642.2
Eclampsia	642.6
Multiple gestation pregnancy	V310, 320, 340, 350, 360, 370
Antepartum hemorrhage	641.3, 641.8, 641.9
Chorioamnionitis	658.4
Premature labor	644.0 (Threatened) and 644.2 (early onset)
Oligohydramnios	658.0
Premature/prolonged rupture of membranes	658.1, 658.2
Use of tobacco, alcohol, or illicit drugs during pregnancy	649.0, 648.3
Maternal medical conditions
Hypertension	642.1
Diabetes mellitus	648.0
Liver disorders	646.7
Congenital heart disease	648.5
Asthma	493.0
Renal disease	646.2
Collagen vascular diseases/lupus	695.4
Presence of major congenital anomaly	740–759.9

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. Selective serotonin reuptake inhibitors for depression in pregnancy
2. Maternofetal pharmacokinetics and fetal lung responses in chronically catheterized sheep receiving constant, low-dose infusions of betamethasone phosphate
3. 8: Detection & implications of lactobacillus iners sub-strain diversity
4. 14: Susceptibility of bacterial vaginosis (BV)-associated bacteria to secnidazole, metronidazole, tinidazole and clindamycin

Stay updated, free articles. Join our Telegram channel

Join