A thought-provoking paper published this month in Health Affairs shows stunning variation in rates of obstetrical complications across U.S. hospitals. This type of research is important and necessary because focusing on averages masks potentially large differences in how patient care is provided and how clinical decisions are made.

From a policy perspective, it’s crucial to identify and learn from hospitals that are “positive deviants,” that is – hospitals with better-than-expected quality of care. From a pregnant woman’s perspective, having information on hospital rates of hemorrhage, infection, or laceration during childbirth is a high priority.

Authors Laurent Glance and colleagues add to a growing literature on variation in hospital-based maternity care. Having useful quality measurement and reporting strategies to guide policy and patient decisions is an essential next step. Indeed, Glance and colleagues conclude by urging clinicians and policymakers to “develop comprehensive quality metrics for obstetrical care and focus on improving obstetrical outcomes.”

This is a laudable and logical conclusion based on their work, but the question is – how?  The data presented by Glance, combined with other emerging evidence, indicate that the development and effective use of quality measurement to improve outcomes in obstetrics may be a long and complicated journey. But, it’s one well worth taking.

Step 1: Reduce Variability in Clinical Processes of Care

Quality improvement focused on standardization of care processes has been the cornerstone of highly-regarded professional efforts, and many hospitals and health care delivery systems have successfully standardized obstetric practices; for example, reducing or eliminating elective deliveries before 39 weeks gestation. However, care processes varied substantially among the 25 tertiary care academic health centers included in a study published last week by William Grobman and colleagues in the American Journal of Obstetrics and Gynecology.

For example, one hospital reported that fewer than 7 percent of obstetrical patients were admitted with a cervical dilation of less than two centimeters; at another hospital, almost half of obstetrical patients were admitted at less than two centimeters. This type of variability was also observed for other care processes: labor induction, labor augmentation, length of active pushing phase, use of vaginal examinations during labor, and use of both regional and general anesthesia in childbirth. Further, this study showed that processes of care – the major focus of current quality metrics in obstetrics – varied widely across hospitals, and did not explain variation in adverse outcomes. Indeed, 50-100 percent of inter-hospital variation was unexplained by patient, hospital, or care process factors.

These vast inter-hospital differences in care patterns pose both a management challenge for clinical obstetrics and a statistical challenge for quality measurement. It’s hard to find a signal amid statistical noise, and the variability in care processes shown by Grobman and colleagues may be one of the reasons for their lack of explanatory power in accounting for variability between hospitals in performance on obstetrical outcomes. Where clinical evidence indicates that care processes improve outcomes, efforts to reduce variability in these care processes should precede measurements of their ability to improve outcomes and explain variation in outcomes across hospitals.

This is already underway in the field of obstetrics. For example, a recent Consensus Opinion from American College of Obstetricians and Gynecologists and the Society for Maternal Fetal Medicine provides specific guidance for clinical practices to safely reduce rates of primary cesarean. Such professional guidelines, if taken seriously and widely adopted within health care delivery systems, can go a long way toward reducing variability in care processes for obstetric patients. Before a performance gap in outcomes can be narrowed, there needs to be greater adherence to evidence-based care processes.

Step 2: Develop a Data Infrastructure for Perinatal Care Quality and Make Information Accessible

In addition to the lack of quality metrics, Glance and colleagues mention the importance of developing a data infrastructure to enable the use and reporting of quality metrics. I couldn’t agree more. Last week, the CDC announced that by 2015, all states will be using the 2003 revision of the U.S. Birth Certificate. While this is certainly laudable, it’s important to move faster, and we can do better in regards to designing data systems that capture the information needed to measure, report on, and improve quality. Most maternity care clinicians don’t know their own cesarean rates. Most payers struggle to understand the obstetric care processes they finance because administrative claims data do not contain information on parity and gestational age, important components of care management for labor and delivery.

A completely integrated, accessible, interoperable perinatal data registry may seem “pie-in-the sky,” but the Medicaid Medical Directors Network and Association of State and Territorial Health Officials are working on systems to reliably link birth certificates with Medicaid claims data. In addition, some states, such as California, Illinois, Louisiana, and Oklahoma are working toward real-time data systems to provide accurate, population-level data to inform policy and program implementation. Medicaid programs pay for almost half of all U.S. births, so the public sector has a strong fiscal interest in improving birth outcomes. Finally, one can look to our northern neighbors in British Columbia for an excellent example of a population-based, provincial perinatal database registry.

Opportunities provided by the Health Information Technology for Economic and Clinical Health (HITECH) Act of 2009 and the Affordable Care Act (ACA) of 2010 may reshape the data landscape in a way that is more favorable to obstetric care quality measurement. But we need not wait for the perfect data infrastructure to start reporting available information publicly. Whether the most relevant reporting unit is the hospital, provider group, or individual clinician remains a topic of considerable debate. But given the strong and growing body of evidence showing hospital-level variations in obstetric care, public reporting of hospital-level data seems a reasonable place to start.

Four million American women will give birth this year in U.S. hospitals; up to 13 percent of these women will leave the hospital with both a newborn baby and severe obstetrical complications, such as a hemorrhage, infection, or laceration. Obstetric complications are not entirely random; skilled clinicians and informed patients in a data-driven, supportive health care system can minimize risk of adverse outcomes; for example, published research shows that comprehensive initiatives can reduce a hospital’s maternal adverse outcomes index. Consistent use of evidence-based care processes should reduce variability in hospital rates of obstetric complications. This is an empirical question, and emerging research suggests a need for a comprehensive answer.