This study aimed to assess the coverage and reliability of the Hospital Admissions Information System (SIH/SUS) in the Municipality of Rio de Janeiro, Brazil. Data for women who gave birth in hospitals belonging to the Unified National Health System (SUS) were compared with data from medical records and interviews from a morbidity and mortality survey. Reliability was almost perfect for age (intraclass correlation coefficient: 0.95) and mode of delivery (kappa = 0.94). The sensitivity and specificity of the SIH/SUS for cesarean section were 98.6% and 99.3%, respectively. Reliability of data on addresses was inversely related to level of data aggregation: 0.76 for programmatic areas and 0.68 for administrative regions. Of 6,835 hospital admissions covered by the SUS, 11.9% showed no records in the SIS-SUS. The high cesarean rate among admissions without prior authorization by the public system appears to be related to regulations setting limits on reimbursement according to mode of delivery.