OBJECTIVE: To assess the trends and patterns of referral from midwives to obstetricians within the Dutch maternity care system from 1988 to 2004, and the differences in referral patterns between nulliparous and parous women. DESIGN: A descriptive study. SETTING: The Dutch midwifery database (LVR1), which monitored 74% (1988) to 94% (2004) of all midwifery care in the Netherlands between 1988 and 2004. POPULATION: A total of 1 977 006 pregnancies, attended by a primary care level midwife. METHODS: The indications for referral from midwifery to obstetric care were classified into fifteen groups (eight antepartum, six intrapartum and one postpartum). The trends in referrals of these indications were analysed by general linear models. MAIN OUTCOME MEASURES: Trends in the percentage of antepartum, intrapartum and postpartum referrals from midwifery care to obstetric care; trends in the specific indications for referral; contribution of different groups of the indications to the trend. RESULTS: From 1988 to 2004 an increase of 14.5% (from 36.9 to 51.4%) occurred in referrals from primary midwifery care to secondary obstetric care either during pregnancy, childbirth or in the postpartum period. The timing of the referrals was as follows: antepartum +9.0%, intrapartum +5.2% and postpartum +0.3%. In parous women, the increase in referrals was greater (+16.6%) than in nulliparous women (+12.3%) (P = 0.001). The commonest indications for referrals in nulliparous women were anticipated or evident complications due to 'failure to progress in the first or second stage' and 'fetal distress'. Parous women were most commonly referred for anticipated or evident complications due to 'medical history' and 'fetal distress'. In nulliparous women, 52% of the increase in referrals was related to the need of pain relief and occurrence of meconium-stained amniotic fluid; in parous women, 54% of the increase in referrals was related to the general medical and obstetrical history of the women, particularly previous caesarean section, and the occurrence of meconium-stained amniotic fluid. CONCLUSIONS: During a 17-year period, there was a continuous increase in the referral rate from midwives to obstetricians. Previous caesarean section, requirement for pain relief and the presence of meconium-stained amniotic fluid were the main contributors to the changes in referral rates. Primary prevention of caesarean section and antenatal preparation for childbirth are important interventions in the maintenance of primary obstetric care for low-risk pregnant women.