Materials and Methods

In the summer of 2007, 27 pilot facilities of the Hospital Corporation of America in 14 states were chosen for an investigation into the frequency of elective delivery at <39 weeks of gestation and the impact of this practice on neonatal outcomes. Facilities were chosen for geographic and demographic representation of our larger system that is responsible for the delivery of approximately 220,000 babies annually in 21 states. Thirteen facilities had annual delivery volumes of <2000; 9 facilities had delivery volumes of 2000-4000, and 5 facilities had delivery volumes of >4000. This system has been shown previously to be roughly representative of the United States as a whole. During a 3-month period, data were collected from >17,000 deliveries.

Based on the observed morbidity that is associated with this early term delivery, we then instituted efforts to reduce its frequency throughout our system. After a period of physician and nursing education that included the provision of published practice guidelines and our own internal data, medical staffs at all hospitals were informed of our intent to restrict this practice on the basis of patient safety considerations. However, medical staffs were allowed to choose 1 of 3 approaches to reduction of this practice: (1) a “hard stop” approach that involved the adoption of a policy that would prohibit purely elective inductions and primary and repeat cesarean deliveries at <39 weeks of gestation. This policy would be enforced by hospital staff members who were empowered to refuse to schedule any such deliveries. Questionable “indications” would be handled in the standard manner by accessing chain of command. (2) A “soft stop” approach that would include adoption of a similar policy to that described earlier. In contrast to the “hard stop” approach, compliance would be left up to individual physicians, and elective deliveries at <39 weeks of gestation would be allowed if ordered by the attending physician. However, all such cases would be referred to the local peer review committee for evaluation and potential action. (3) An “education only” approach that would involve the provision of available literature to attending physicians and both internal and professional association recommendations against this practice, which was also provided with the first 2 approaches. However, no formal policy prohibiting this practice would be adopted by the medical staff.

Data regarding physician compliance and neonatal outcomes were collected exactly 2 years later (2009) during the same 3 months of the year (May, June, July) and compared with the baseline data from these same 27 facilities in 2007. Analysis of identical facilities during identical months of the year within a 2-year period was necessary to minimize confounding effects of changes in patient or provider population or of scheduling concerns. Because of a concern regarding potential development of “creative” indications by staff physicians, we tracked rates of each type of planned delivery (elective and indicated) during these 2 time periods as an internal control. A planned delivery was defined as 1 in which the mother delivered after entering the labor and delivery suite not in labor and with intact membranes. An elective delivery was defined as a planned delivery without a recognizable medical or obstetric indication for delivery by either the attending physician or the nurse who collected the data. This included inductions and primary and repeat cesarean deliveries. Gestational age was assigned based on the best estimate of the attending clinician according to both menstrual history and prenatal sonography. For the overall reduction in rates of elective early term delivery and newborn intensive care unit admissions, the unit of analysis was the individual delivery.

For the comparison of departmental policy, facility rates were used as the unit of analysis. Statistical analysis for the overall performance and neonatal outcome data was performed with the χ ² test with Yates correlation correction. One-way analysis of variance and Friedman repeated measures analysis of variance on ranks with all pairwise multiple comparison procedures (Student-Newman-Keuls method) and 2-way analysis of variance with multiple comparisons vs control group (Holm-Sidak method) were used to compare sequential performance differences in the 3 study groups. Significance was set at a probability value of .05. This was a quality improvement project that used deidentified data for analysis. Exemption from institutional review board review was obtained based on 45CFR46.101(b) and 46.102(f) and 45CFR164.514(a)-(c) of the Health Insurance Portability and Accountability Act. However, institutional review board approval had been obtained for the control data publication.