The criticality of referral makes it imperative to study its patterns and factors influencing it at a health systems level. This study of referral in Lagos, Nigeria is based on health records of 4181 pregnant women who presented with obstetric emergencies at one of the 24 comprehensive emergency obstetric care (EmOC) facilities in the state between November 2018 and October 2019 complemented with distance and time data extracted from Google Maps. Univariate, bivariate, and multivariate analyses were conducted. About a quarter of pregnant women who presented with obstetric emergencies were referred. Most referrals were from primary health centres (41.9 %), private (23.5 %) and public (16.2 %) hospitals. Apart from the expected low-level to high-level referral pattern, there were other patterns observed including non-formal, multiple, and post-delivery referrals. Travel time and distance to facilities that could provide needed care increased two-fold on account of referrals compared to scenarios of going directly to the final facility, mostly travelling to these facilities by private cars/taxis (72.8 %). Prolonged/obstructed labour was the commonest obstetric indication for referral, with majority of referred pregnant women delivered via caesarean section (52.9 %). After adjustment, being married, not being registered for antenatal care at facility of care, presenting at night or with a foetus in distress increased the odds of referral. However, parity, presentation in the months following the commissioning of a new comprehensive EmOC facility or with abortion reduced the likelihood of being referred. Our findings underscore the need for health systems strengthening interventions that support women during referral and the importance of antenatal care and early booking to aid identification of potential pregnancy complications whilst establishing robust birth preparedness plans that can minimise the need for referral in the event of emergencies. Indeed, there are context-specific influences that need to be addressed if effective referral systems are to be designed.