BACKGROUND: An estimated 2.6 million stillbirths occur every year, with the majority occurring in low- and middle-income countries. Understanding the cause of and factors associated with stillbirth is important to help inform the design and implementation of interventions aimed at reducing preventable stillbirths. METHODS: Population-based surveillance with identification of all stillbirths that occurred either at home or in a health facility was introduced in four districts in Bangladesh. Verbal autopsy was conducted for every fifth stillbirth using a structured questionnaire. A hierarchical model was used to assign likely cause of stillbirth. RESULTS: Six thousand three hundred thirty-three stillbirths were identified for which 1327 verbal autopsies were conducted. 63.9% were intrapartum stillbirths. The population-based stillbirth rate obtained was 20.4 per 1000 births; 53.9% of all stillbirths occurred at home. 69.6% of mothers had accessed health care in the period leading up to the stillbirth. 48.1% had received care from a highly trained healthcare provider. The three most frequent causes of stillbirth were maternal hypertension or eclampsia (15.2%), antepartum haemorrhage (13.7%) and maternal infections (8.9%). Up to 11.3% of intrapartum stillbirths were caused by hypoxia. However, it was not possible to identify a cause of death with reasonable certainty using information obtained via verbal autopsy in 51.9% of stillbirths. CONCLUSIONS: Introducing surveillance for stillbirths at community level is possible. However, verbal autopsy yields limited data, and the questionnaire used for this needs to be revised and/or combined with information obtained through case note review. Most women accessed and received care from a qualified healthcare provider. To reduce the number of preventable stillbirths, the quality of antenatal and intrapartum care needs to be improved.