Obstetrical delivery is the most frequent cause of hospital admissions (4 million) in the US. OBJECTIVE: To assess the relationship of patient and hospital-related characteristics with maternity length of stay (LOS) and charges for obstetrical deliveries. METHODS: Records of 62,003 obstetrical deliveries were extracted from a Nationwide Inpatient Sample of the Healthcare Cost and Utilization Project data (DRG 371 for cesarean delivery without complications and 373 for vaginal delivery without complications). Regression analyses were conducted to predict maternity LOS and total charges based on type of delivery, age, race, number of diagnoses at discharge, primary payer (Medicaid, private insurance/HMO, self-pay, other), and hospital characteristics of bed size, region (northeast, midwest, south, west), location (rural, urban), teaching status, and control/ownership. RESULTS: In 1997, 82% of the deliveries were vaginal while 18% were cesarean. Average LOS for vaginal and cesarean deliveries was 1.86 and 3.36 days, respectively. Mean total charges for vaginal deliveries were $3,778 as compared to $7,243 for cesarean deliveries. Regression results indicate that type of delivery was the strongest predictor of LOS and total charges. Women with a cesarean delivery stayed on an average 1.3 days more in the hospital and incurred 64% more charges than women with vaginal delivery. For vaginal delivery, an additional day of hospitalization increased total charges by 12.4% as compared to 6.7% for cesarean delivery. Medicaid patients had shorter LOS than private insurance/HMOs, and longer LOS than self-paying patients. Northeast hospitals had greater LOS while hospitals in the West had higher charges. Urban hospitals and teaching hospitals had higher LOS and charges. CONCLUSIONS: Significant variations in maternity LOS and total charges were observed on the basis of type of delivery and hospital-related characteristics, which warrant further investigation. Additionally, the impact of LOS and total charges on mother and infant morbidity should be explored.