Objective To identify factors that are predictive of late initiation of antenatal care in England and Wales.Design A multivariate binomial regression model was constructed to examine the association between clinical, provider and sociodemographic characteristics and late initiation of antenatal care. Setting Nine maternity units in Northern England and North Wales.Population A total of 20,771 women with a singleton pregnancy who delivered a liveborn or stillborn baby between 1 August 1994 and 31 July 1995. All analyses were based on the 17,765 (85.5%) women for whom information on gestational age at initial presentation for antenatal care and other variables incorporated into the regression model was retrievable from the case records. Results Primiparous women of high obstetric risk were 13.4% more likely to initiate antenatal care after 10 weeks of gestation than a low risk reference group (adjusted OR 1.134, 95% CI 1.011, 1.272; P ¼ 0.0312), and 34.3% more likely to initiate antenatal care after 18 weeks of gestation (adjusted OR 1.343, 95% CI 1.046, 1.724; P ¼ 0.0208). This association between high obstetric risk status and late initiation of antenatal care was not replicated among multiparous women. When the effects of other independent variables on gestational age at booking were examined, the following characteristics were associated with failure to initiate antenatal care by 10 weeks of gestation ( P 0.05): maternal age at booking, smoking status, ethnicity, type of hospital at booking, the planned pattern of antenatal care and the planned place of delivery. Adopting a criterion of 18 weeks of gestation exacerbated the association between clinical and sociodemographic characteristics and late initiation of antenatal care, but appeared to dilute the association between provider characteristics and late initiation of antenatal care. Conclusions There is a pressing need for further research to identify the specific concerns of late bookers, to identify areas where new interventions might encourage the uptake of services and to gauge the likely impact of increased dissemination of information about the availability of antenatal care services.