Objective. The aim of this study was to analyse infant mortality among infants born extremely preterm in relation to mode of delivery, maternal diagnosis, and different institutional policies. Methods. We conducted a national tertiary health care center study using Swedish Medical Birth Register (MBR) data from 1990 to 2002, to examine the 2,094 live births of infants at 23+0 to 27+6 weeks gestation. We assessed the association between mode of delivery, gestational age (GA), calendar year, maternal condition, and institutional policies on infant mortality outcome. Results. At 23–25 weeks, 38% of infants (range: 34–69%) were delivered by cesarean section (CS), while at 26–27 weeks, 66% (59–80%) were delivered by CS. The CS rate for fetal or maternal indications was 98% in cases of pre-eclampsia/eclampsia, 42% for premature rupture of membranes (PROM), 68% for hemorrhage, 76% for PROM+hemorrhage, 56% for breech presentation, and 30% for preterm vertex with no other complications. After cases of pre-eclampsia/eclampsia were excluded, vaginal delivery was associated with a small increase of risk for infant death. Vaginal delivery was associated with a significantly increased risk for infant death in breech presentations and multiple births, while vaginal delivery posed a non-significant risk increase for PROM and hemorrhage. For preterm vertex without any other complications, 4 out of 5 infants were delivered vaginally without any risk increase. Conclusion. This study reports high CS rates for very preterm births at Swedish hospitals. In performing CS for very preterm infants, this study suggests a survival advantage for certain maternal conditions, but not for preterm labor with a vertex presentation without other obstetrical complications.