Childbirth outcomes in the United States vary tremendously between hosiptals.1–3 Hospital-level variation in cesarean delivery rates is particularly stark with 10-fold variation from 7 to 70%.4 Cesarean deliveries are associated with increased rates of severe morbidity, longer hospitalizations, and greater average costs relative to vaginal deliveries, and approximately 45% of these procedures may be avoidable.5–8 Hospital-level variation in cesarean delivery rates persists after adjusting for maternal clinical and sociodemographic factors suggesting a significant opportunity to improve the safety, affordability, and experience of care.4,9–11
In nearly every industry, interorganizational variation in performance can be partly explained by differences in management.12,13 In health care, better hospital management has been associated with improved quality of care and increased adoption of clinical best practices.14–17 In obstetrics, “systems factors” such as patient volume have been shown to affect quality of care.18,19 Nonetheless, the specific management practices of labor and delivery units such as methods for adjusting nurse staffing levels to accommodate changes in patient volume have not been well assessed.
Labor and delivery units pose exceptional management challenges. They must care for patients with a wide spectrum of risks and rapidly shifting needs despite low financial operating margins.20 Given this combination of uncertainty and scarcity, we hypothesized that quality of care may be sensitive to operational pressures to deliver neonates more efficiently. In this article, we assess whether labor and delivery unit management practices that proactively mitigate unit challenges are independently associated with a lower risk of primary cesarean delivery, prolonged length of stay, and adverse maternal outcomes in low-risk patients.
MATERIALS AND METHODS
There are few valid measurement strategies for evaluating heterogeneous management practices across diverse institutions. Working with an interdisciplinary team of clinicians and management experts, we developed an assessment instrument based on the World Management Survey method, which uses telephone interviews to capture a wide scope of management complexity while minimizing the time burden on participating managers.21 The telephone interviews collect qualitative practice descriptions, which are then quantitatively scored using Likert-style scales to allow for comparison across institutions in a standard fashion.
We generated an initial list of nonclinical systems factors and management strategies that may affect cesarean delivery rates through a literature review and interviews with 18 experts who had direct responsibility for managing or supporting the management of labor and delivery units. We organized the list of factors based on the types of management strategies involved. For example, coordination among anesthesia, neonatology, and other services was examined collectively in terms of team collaboration and communication. Then, we engaged a scientific advisory board of national thought leaders representing stakeholders across obstetrics and nursing management, operations management, and health care quality measurement. The advisory board selected 16 modifiable management factors through a modified Delphi consensus process. During the modified Delphi process, the board members first independently completed a questionnaire to elicit their perceptions of which management factors are most likely to affect maternal outcomes. Then, during an in-person board meeting, we reviewed these responses collectively. An experienced neutral facilitator helped reconcile areas of disagreement among board members (Fig. 1).
Based on the World Management Survey model, the instrument consists of two, structured 50-minute telephone interviews that are conducted with the primary nurse manager and the primary physician manager of the labor and delivery unit at each hospital. The interview with the nurse manager covers 11 management factors and the interview with the physician manager covers 10 management factors. Five management factors in areas of overlapping expertise are assessed in both the nurse and physician interviews to allow for within-hospital comparisons. Based on information collected from our expert interviews, we developed Likert-style scoring scales for each of the management factors. Hospitals with the lowest management scores had “reactive” management practices that only address management problems as they occur (eg, no process for tracking or anticipating bottlenecks in patient flow), whereas hospitals with the highest management scores had “proactive” management practices that preemptively mitigate unit challenges (eg, a well-functioning process to identify and address bottlenecks in patient flow).21
We used telephone and email solicitations to recruit 53 hospitals that are members of the National Perinatal Information Center, a nonprofit data analysis and reporting organization that provides quality assurance and benchmarking services. From February to March 2015, a single interviewer conducted and scored interviews with at least one nurse manager and at least one physician manager at each of 11 pilot hospitals. For 7 of the 11 pilot hospitals, the interviewer conducted interviews with one or two additional managers in the same clinical service line to confirm an appropriate match between the instrument content and department-level manager expertise. Only data from the primary nurse and physician manager interviews from the pilot sites were included in the final analyses to ensure comparability across hospitals. All interviews were audio-recorded.
As a validation step, we conducted in-person, half-day site visits at all 11 pilot hospitals to confirm the accuracy of the information collected through their telephone interviews. During each visit, we met with the previously interviewed managers to ensure we were interpreting responses correctly and toured each labor and delivery unit to directly observe management practices in context.
From August to December 2015, the same interviewer conducted and scored interviews with the primary department-level nurse manager and physician manager at each of the remaining hospitals. The interviewer was blinded to all outcomes for these 42 hospitals. Multiple domain and methodology experts, including the interviewer, independently scored interview responses using the scoring scales and then collaboratively edited the scales over multiple rounds of scoring to improve reliability and usability. Two additional blinded researchers scored a random subsample of 10 interviews. We calculated a weighted κ coefficient from their scores to quantify instrument reliability.
For the five management factors that were included in both nurse and physician manager interviews, managers from the same hospital received scores within 2 points of each other in 87% of cases. To reconcile discrepant scores, we performed a sensitivity analysis that showed similar results using either nurse manager or physician manager scores for the overlapping factors. Based on these similarities, we report the results using physician scores in the tables rather than averaging them to preserve measurement precision and interpretability.
Our primary outcome was low-risk primary cesarean delivery with low-risk defined as term, singleton, vertex pregnancies (Agency for Healthcare Research and Quality Inpatient Quality Indicators #33).22 Our secondary outcomes were prolonged length of stay during the patients' childbirth hospitalization and four measures of adverse outcomes. We defined prolonged length of stay for mothers as greater than 3 days for vaginal deliveries and 5 days for cesarean deliveries.23–25 Our four measures of adverse outcomes were 1) experiencing at least one indicator of severe maternal morbidity as defined by the Centers for Disease Control and Prevention, 2) maternal obstetric infection, 3) postpartum hemorrhage (International Classification of Diseases, 9th Revision, Clinical Modification code 666), and 4) blood transfusion (Appendix 1, available online at http://links.lww.com/AOG/A971).26,27
For the outcomes analyses, we excluded patients with diagnosis codes related to antepartum care (International Classification of Diseases, 9th Revision, Clinical Modification codes 630–641, 644) and patients with missing demographic data in their administrative claims record.28 Two hospitals did not report race and ethnicity for any patients and were therefore not included in the final outcomes analyses given the large potential effect of racial and ethnic disparities on our primary and secondary outcomes.29 The excluded population with missing data was not significantly different from the included population with complete cases (Appendix 2, available online at http://links.lww.com/AOG/A971).
We analyzed management data in the context of patient characteristics that were derived from the most recently available administrative claims data (2013–2014) and hospital characteristics that were either derived from the administrative claims data or reported in routine National Perinatal Information Center member surveys. We performed all statistical analyses in Stata 14.0 and SAS 9.4. We collected and managed all management scores using REDCap electronic data capture tools hosted at the Harvard T. H. Chan School of Public Health.30
To assess whether there was significant variation across hospitals for each of the 16 management factors, we used two-sided χ2 tests of variance.31 To assess variation within hospitals, we calculated Pearson correlations between individual management factors with a Bonferroni correction for multiple comparisons.
Given the variation in management scores we observed across and within hospitals for all 16 management factors, we performed an exploratory factor analysis to ascertain management themes. These management themes represent patterns among multiple management factors that tend to trend together while accounting for the possibility of interactions between management factors. Management factors with loading values greater than 0.4 were strongly aligned with a given management theme and incorporated into a composite score for that management theme.
We analyzed pairwise associations between management themes and hospital and patient characteristics using Pearson correlations or canonical correlations.
We constructed patient-level relative risk (RR) models to assess the associations between management themes and maternal outcomes. The patient-level models treated labor and delivery unit management as an independent risk factor for any given outcome with equal weight as a medical comorbidity such as diabetes. We controlled for patient-level covariates, including maternal age, maternal age-squared, maternal race, gestational diabetes, hypertension, and private insurance, and hospital-level covariates, including percentage of privately insured patients, teaching service, midwifery service, total delivery volume, neonatal intensive care unit level, and geographic region. We additionally controlled for inductions of labor in the model for prolonged length of stay.
The RR models used a log link function and were estimated within a generalized estimating equations framework to account for hospital-level clustering.32 The P values and CIs were adjusted for multiple testing using a Bonferroni stepdown procedure.33 To calculate the potential magnitude the association between management and maternal outcomes, we compared the lowest and highest scoring hospitals for each management theme.
The Harvard Human Resource Protection Program's institutional review board approved the study protocol and consent processes. All enrolled hospitals signed written participation agreements and each manager directly interviewed gave verbal consent to participate at the beginning of their interview.
The final management assessment instrument included an interview guide and scoring scale for each of the 16 selected management factors (Table 1; Appendix 3, available online at http://links.lww.com/AOG/A971). The instrument had a strong weighted κ score for interrater reliability (κ=0.92, P<.01).
The 226,463 patients included in our sample had a mean age of 28.6±5.8 years and were 55.7% white, 20.6% black, 7.6% Hispanic, and 16.1% other races. The 51 hospitals included in the final outcome analyses represent a diverse range of delivery volume, health care provider mix, and geography, but on average skew toward larger annual delivery volumes and academic medical centers with higher acuity patients than the national averages (Table 2).34 The nurse managers interviewed had a median of 3 years of experience in their current management role and a median of 14 years of experience at their hospital, and the physician managers interviewed had a median of 2 years of experience in their current management role and a median of 15 years of experience at their hospital.
Across hospitals, χ2 tests of variance showed significant variation on all 16 management factors (P<.01). Within hospitals, only two pairs of factors showed strong, significant, positive correlations: obstetrician availability and communication and coordination (r=0.50, P<.01) and team collaboration and communication and coordination (r=0.68, P<.01).
We excluded performance reporting from the exploratory factor analysis as a result of the magnitude of its negative correlation to the total sum hospital score, which indicated that this factor was unrelated to any of the identified themes. We similarly excluded standardization of processes from the final analysis as a result of weak factor loading (less than 0.4) across all identified themes. The resulting factor analysis identified three themes from the 14 remaining management factors (Table 3). We named the three themes: 1) unit culture management, 2) patient flow management, and 3) nursing management based on the content of the management factors they included. A sensitivity analysis with nurse manager scores found that in cases in which the nurse manager and physician manager scores were discrepant, selecting either score yielded similar results attributed to the relatively small size of the discrepancies. Results are reported using physician manager scores in Table 3.
Hospitals with higher scores in the theme of unit culture management tended to care for older patients (r=0.38, P=.01) and were more likely to be teaching hospitals (r=0.28, P=.05) and located in the Northeast region (r=0.63, P<.01). Hospitals with higher scores in the theme of patient flow management were more likely to care for older patients (r=0.31, P=.03) and privately insured patients (r=0.33, P=.02) and were more likely to have a higher delivery volume (r=0.54, P<.01) and a neonatal intensive care unit level greater than level I (r=0.40, P=.03). There were no significant associations between hospital scores in the theme of nursing management and patient or hospital characteristics.
Patients who received care at the hospital with the highest unit culture management score had a 30% higher RR of low-risk primary cesarean delivery (RR 1.30, 95% CI 1.02–1.66, P=.04) than patients who received care at the hospital with the lowest unit culture management score. There was no significant association between a patient's risk of low-risk primary cesarean delivery and the patient flow management score of the hospital where she delivered. Patients who received care at the hospital with the highest nursing management score had a 47% higher RR of low-risk primary cesarean delivery (RR 1.47, 95% CI 1.13–1.92, P<.01) than patients who received care at the hospital with the lowest nursing management score (Table 4). These risks mean that at a hospital with the average primary cesarean delivery rate in the United States (25.8%), higher unit culture management scores could be associated with an increase in primary cesarean delivery rate as large as 7.7% points (33.5%) and higher nursing management scores could be associated with an increase in primary cesarean delivery rate as large as 12.1% points (37.9%).35
Patients who received care at the hospital with the highest unit culture management score had a 313% higher RR of prolonged length of stay (RR 4.13, 95% CI 1.98–8.64, P<.01) than patients who received care at the hospital with the lowest unit culture management score. Patients who received care at the hospital with the highest patient flow management score had a 77% lower RR of prolonged length of stay (RR 0.23, 95% CI 0.12–0.46, P<.01) than patients who received care at the hospital with the lowest patient flow management score, and patients who received care at the hospital with the highest nursing management score had a 73% lower RR of prolonged length of stay (RR 0.27, 95% CI 0.11–0.62, P<.01) than patients who received care at the hospital with the lowest unit culture management score.
There were no significant associations between a patient's risk of experiencing at least one indicator for severe maternal morbidity or infection and any labor and delivery unit management theme scores. Patients who received care at the hospital with the highest unit culture management score had a 157% higher RR of postpartum hemorrhage (RR 2.57, 95% CI 1.58–4.18, P<.01) and an 87% higher RR of blood transfusion (RR 1.87, 95% CI 1.12–3.13, P=.02) than patients who received care at the hospital with the lowest unit culture management score. There were no significant associations between a patient's risk of experiencing postpartum hemorrhage and blood transfusion and the patient flow or nursing management scores of the hospital where she delivered.
Labor and delivery unit management varied dramatically across and within hospitals in the United States. Unit management was independently related to maternal outcomes, including low-risk primary cesarean delivery, postpartum hemorrhage, blood transfusion, and prolonged length of stay. Higher unit culture and nursing management scores were associated with a higher RR of primary cesarean delivery in low-risk patients. Hospitals with higher cesarean delivery rates may have instituted more proactive management practices to lower their rate leading to a reverse causality effect, where cesarean delivery rates drive management practices rather than vice versa. Further research into this counterintuitive finding should also investigate whether there is a relationship between management practices and management goals. Managers may face competing management goals that do not optimize cesarean delivery rates such as enhancing financial performance. Data to ascertain whether there were additional management goals were not available for this study.
Higher unit culture management scores were also independently associated with a higher RR of a patient experiencing postpartum hemorrhage, blood transfusion, and prolonged length of stay. Hospitals that had higher scores in unit culture management were more likely to care for older patients and more likely to be academic medical centers. Although we control for both age and common comorbidities in our model, we were limited in our ability to fully assess patient complexity at these centers by our use of administrative claims data. For example, the claims-based measure we used to calculate primary cesarean delivery rates among low-risk women does not take into account parity.
In contrast, higher scores in patient flow and nursing management, themes that include proactive adjustment of unit resources, were associated with a lower RR of prolonged length of stay for low-risk patients, but not associated with any measured adverse outcomes. These findings validate the intention of our assessment instrument to measure key elements of labor and delivery unit management that affect the efficiency of unit operations without sacrificing patient safety. Effectively and efficiently allocating scarce unit resources across patients on the unit represents a key priority for managers. The lack of a relationship between these management themes and any measured adverse outcomes indicates a potential opportunity to safely increase operational efficiency.
Some of our measurements may be within the zone of potential bias and must be interpreted in the context of our sample and study design.36 The hospitals studied represented a diverse range of characteristics, but 76% had delivery volumes greater than 2,475 deliveries per year (top 15% in the United States) and 90% had primary cesarean delivery rates below the federal HealthyPeople 2020 target of 23.9%.34,35 These differences may have led to an underestimation of the effect of management on outcomes, particularly cesarean delivery rates.
Furthermore, within our cohort, larger hospitals and academic hospitals tended to have higher management scores. These hospitals tend to care for higher acuity patients and therefore a greater range of acuity overall. Management may be optimized to care for the highest risk patients rather than lower risk patients within these settings. This observation highlights the importance of future work to develop and scale management best practices for smaller, community-based or critical access hospitals.
Our management assessment approach was designed for feasibility and therefore had time-limited constraints. The 50-minute-long telephone interview format and limit of two department-level manager interviewees per hospital also limited the number of management factors we could study as well as the depth and context we could collect. Assuming other management practices would follow the similar pattern of substantial variation within and across hospitals, we would expect to see even greater variation in management in a study that included a larger set of management factors. Interviewing two managers at each site also led to the need to resolve discrepant scores between nurses and physician managers. Although we used physician scores in our reporting of patient outcomes, we found that substituting the nursing scores in our factor analysis did not significantly affect the loading of the management themes we identified.
Despite these limitations, our study provides important insights into the independent relationships between management and maternal outcomes. In an era when hospitals are increasingly accountable for quality of care, this work forms the foundation for future research into how to measure and modify management practices to improve obstetric outcomes.
1. Glance LG, Dick AW, Glantz JC, Wissler RN, Qian F, Marroquin BM, et al. Rates of major obstetrical complications vary almost fivefold among US hospitals. Health Aff (Millwood) 2014;33:1330–6.
2. Friedman AM, Ananth CV, Prendergast E, D'Alton ME, Wright JD. Variation in and factors associated with use of episiotomy. JAMA 2015;313:197–9.
3. Bailit JL, Grobman WA, Rice MM, Spong CY, Wapner RJ, Varner MW, et al. Risk-adjusted models for adverse obstetric outcomes and variation in risk-adjusted outcomes across hospitals. Am J Obstet Gynecol 2013;209:446.e1–30.
4. Kozhimannil KB, Law MR, Virnig BA. Cesarean delivery rates vary tenfold among US hospitals; reducing variation may address quality and cost issues. Health Aff (Millwood) 2013;32:527–35.
5. Witt WP, Wisk LE, Cheng ER, Mandell K, Chatterjee D, Wakeel F, et al. Determinants of cesarean delivery in the US: a lifecourse approach. Matern Child Health J 2015;19:84–93.
6. Liu S, Liston RM, Joseph KS, Heaman M, Sauve R, Kramer MS, et al. Maternal mortality and severe morbidity associated with low-risk planned cesarean delivery versus planned vaginal delivery at term. CMAJ 2007;176:455–60.
7. Deneux-Tharaux C, Carmona E, Bouvier-Colle M-H, Bréart G. Postpartum maternal mortality and cesarean delivery. Obstet Gynecol 2006;108:541–8.
8. Truven Health Analytics. The cost of having a baby in the United States. Ann Arbor (MI): Truven Health Analytics; 2013.
9. Cáceres IA, Arcaya M, Declercq E, Belanoff CM, Janakiraman V, Cohen B, et al. Hospital differences in cesarean deliveries in Massachusetts (US) 2004–2006: the case against case-mix artifact. PLoS One 2013;8:e57817.
10. Kozhimannil KB, Arcaya MC, Subramanian SV. Maternal clinical diagnoses and hospital variation in the risk of cesarean delivery: analyses of a National US Hospital Discharge Database. PLoS Med 2014;11:e1001745.
11. Baicker K, Buckles KS, Chandra A. Geographic variation in the appropriate use of cesarean delivery. Health Aff (Millwood) 2006;25:w355–67.
12. Bloom N, Lemos R, Sadun R, Scur D, Van Reenen J. The new empirical economics of management. J Eur Econ Assoc 2014;12:835–76.
13. Bloom N, Lemos R, Sadun R, Van Reenen J. Does management matter in schools. Econ J (R Econ Soc) 2015;125:647–74.
14. McConnell KJ, Lindrooth RC, Wholey DR, Maddox TM, Bloom N. Modern management practices and hospital admissions. Health Econ 2016;25:470–85.
15. McConnell KJ, Lindrooth RC, Wholey DR, Maddox TM, Bloom N. Management practices and the quality of care in cardiac units. JAMA Intern Med 2013;173:684–92.
16. McConnell KJ, Hoffman KA, Quanbeck A, McCarty D. Management practices in substance abuse treatment programs. J Subst Abuse Treat 2009;37:79–89.
17. Dorgan S, Layton D, Bloom N, Homkes R, Sadun R, Van Reenen J. Management in healthcare: why good practice really matters. London (United Kingdom): McKinsey&Company; 2010.
18. Snowden JM, Darney BG, Cheng YW, McConnell KJ, Caughey AB. Systems factors in obstetric care: the role of daily obstetric volume. Obstet Gynecol 2013;122:851–7.
19. Clark SL, Perlin JB, Fraker S, Bush J, Meyers JA, Frye DR, et al. Association of obstetric intervention with temporal patterns of childbirth. Obstet Gynecol 2014;124:873–80.
20. Hung P, Kozhimannil KB, Casey MM, Moscovice IS. Why are obstetric units in rural hospitals closing their doors? Health Serv Res 2016;51:1546–60.
21. Bloom N, Van Reenen J. Measuring and explaining management practices across firms and countries. Q J Econ 2007;122:1351–408.
22. Agency for Healthcare Research and Quality. Inpatient quality indicator #33: primary cesarean delivery rate, uncomplicated. Rockville (MD): Agency for Healthcare Research and Quality; 2009.
23. Kozhimannil KB, Shippee TP, Adegoke O, Virnig BA. Trends in hospital-based childbirth care: the role of health insurance. Am J Manag Care 2013;19:e125–32.
24. Snowden JM, Cheng YW, Emeis CL, Caughey AB. The impact of hospital obstetric volume on maternal outcomes in term, non-low-birthweight pregnancies. Am J Obstet Gynecol 2015;212:380.e1–9.
25. Torvie AJ, Callegari LS, Schiff MA, Debiec KE. Labor and delivery outcomes among young adolescents. Am J Obstet Gynecol 2015;213:95.e1–8.
27. Goff SL, Pekow PS, Avrunin J, Lagu T, Markenson G, Lindenauer PK. Patterns of obstetric infection rates in a large sample of US hospitals. Am J Obstet Gynecol 2013;208:456.e1–13.
28. Wanderer JP, Phil M, Leffert LR, Kuklina EV, Callaghan WM, Bateman BT. Epidemiology of obstetric-related ICU admissions in Maryland: 1999–2008. Crit Care Med 2013;41:1844–52.
29. Bryant AS, Worjoloh A, Caughey AB, Washington AE. Racial/ethnic disparities in obstetrical outcomes and care: prevalence and determinants. Am J Obstet Gynecol 2010;202:335–43.
30. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform 2009;42:377–81.
31. Snedecor GW, Cochran WG. Statistical methods. 8th ed. Iowa City (IA): Iowa State University Press; 1989.
32. Carter RE, Lipsitz SR, Tilley BC. Quasi-likelihood estimation for relative risk regression models. Biostatistics 2005;6:39–44.
33. Holm SA Simple sequentially rejective multiple test procedure. Scand J Stat 1979;6:65–70.
34. Simpson KR. An overview of distribution of births in United States hospitals in 2008 with implications for small volume perinatal units in rural hospitals. J Obstet Gynecol Neonatal Nurs 2011;40:432–9.
36. Grimes DA, Schulz KF. False alarms and pseudo-epidemics. Obstet Gynecol 2012;120:920–7.