This study investigates the relationship between adverse pregnancy outcomes in high-risk African American women in Washington, DC and sociodemographic risk factors, behavioral risk factors, and the most common and interrelated medical conditions occurring during pregnancy: diabetes, hypertension, preeclampsia, and Body Mass Index (BMI). Data are from a randomized controlled trial conducted in 6 prenatal clinics. Women in their 1st or 2nd trimester were screened for behavioral risks (smoking, environmental tobacco smoke exposure, depression, and intimate partner violence) and demographic eligibility. 1,044 were eligible, interviewed and followed through their pregnancies. Classification and Regression Trees (CART) methodology was used to: (1) explore the relationship between medical and behavioral risks (reported at enrollment), sociodemographic factors and pregnancy outcomes; (2) identify the relative importance of various predictors of adverse pregnancy outcomes; and (3) characterize women at the highest risk of poor pregnancy outcomes. The strongest predictors of poor outcomes were prepregnancy BMI, preconceptional diabetes, employment status, intimate partner violence, and depression. In CART analysis, preeclampsia was the first splitter for low birthweight; preconceptional diabetes was the first splitter for preterm birth (PTB) and neonatal intensive care admission; BMI was the first splitter for very PTB, large for gestational age, Cesarean section and perinatal death; employment was the first splitter for miscarriage. Preconceptional factors strongly influence pregnancy outcomes. For many of these women, the high risks they brought into pregnancy were more likely to impact their pregnancy outcomes than events during pregnancy.
Trial registration: ClinicalTrials.gov NCT00381823.