Among the most critical decisions faced by pregnant women and their providers is whether to intervene to deliver or allow the pregnancy to continue, trying to optimize the outcome for mothers and infants. Failure to intervene by labor induction or prelabor Cesarean delivery can lead to adverse health consequences for the mother and infant, whereas intervention results in early delivery with attendant health concerns. We propose to conduct an observational study of a large population of pregnancies eligible for early delivery intervention using a design and data analysis approach that effectively simulates a series of week-by-week intervention trials. We propose to evaluate the risks and benefits of intervention starting with records from 96,000 singleton deliveries that occurred at Women and Infants hospital in Rhode Island over the period 2002-2012. We will restrict the cohort to the ~13,500 pregnancies with one of the three most common potential indications for intervention: fetal growth restriction, maternal diabetic disorders, and maternal hypertensive disorders. For each of these pregnancies, we will abstract key clinical data relevant to the ongoing decision regarding intervention (timing, severity, etc.), with preliminary identification based on administrative data on diagnoses followed by review of prenatal and inpatient medical records to confirm diagnoses and identify determinants of intervention. For each week of each patient's pregnancy between 34 and 38 weeks, we will determine whether they had the condition that put them at risk of intervention for delivery and whether or not the intervention was, in fact, done. The consequences of that weeks' decision will be examined, as will the decisions made in subsequent weeks for those still at risk of intervention. Key infant outcomes to be evaluated are admission to the neonatal intensive care unit, adverse respiratory outcomes, length of hospital stay, and an index of neonatal morbidity; for the mothers, we will assess unplanned Cesarean delivery and duration of hospitalization. The analysis will use propensity scores with multiple imputation to equalize covariates at a given gestational age for those who did and did not receive interventions in each week of gestation from 34 to 38, as well as for the preterm and early term period overall. Subject to the inherent limitations resulting from non-randomization, this approach will closely approximate a trial by accounting for the clinical details of each pregnancy to control for confounding by indication and considering the longitudinal nature of the intervention decisions and their health consequences. Insights gained will have direct relevance to clinical decision-making regarding frequently encountered complications of pregnancy.
Decisions regarding intervention for delivery are based on the clinician's and patient's judgments, balancing the risks and benefits for the infant and mother. At present, there is very limited scientific evidence to guide such judgments. We propose to conduct a study that will assess the consequences of those week-by-week decisions from 34-38 weeks' gestation among pregnancies at risk of such interventions, providing critical information to inform clinical practice and encourage changes, where needed, in the threshold for intervention
