Objective: To test the hypothesis that it is possible to select a group of low risk women who can start labour in midwife-led care without having increased rates of severe adverse maternal outcomes compared to women who start labour in secondary care. Design and Methods: We conducted a nationwide cohort study in the Netherlands, using data from 223 739 women with a singleton pregnancy between 37 and 42 weeks gestation without a previous caesarean section, with spontaneous onset of labour and a child in cephalic presentation. Information on all cases of severe acute maternal morbidity collected by the national study into ethnic determinants of maternal morbidity in the Netherlands (LEMMoN study), 1 August 2004 to 1 August 2006, was merged with data from the Netherlands Perinatal Registry of all births occurring during the same period. Our primary outcome was severe acute maternal morbidity (SAMM, i.e. admission to an intensive care unit, uterine rupture, eclampsia or severe HELLP, major obstetric haemorrhage, and other serious events). Secondary outcomes were postpartum haemorrhage and manual removal of placenta. Results: Nulliparous and parous women who started labour in midwife-led care had lower rates of SAMM, postpartum haemorrhage and manual removal of placenta compared to women who started labour in secondary care. For SAMM the adjusted odds ratio's and 95% confidence intervals were for nulliparous women: 0.57 (0.45 to 0.71) and for parous women 0.47 (0.36 to 0.62). Conclusions: Our results suggest that it is possible to identify a group of women at low risk of obstetric complications who may benefit from midwife-led care. Women can be reassured that we found no evidence that midwife-led care at the onset of labour is unsafe for women in a maternity care system with a well developed risk selection and referral system.