Background: Encouraging institutional birth is an important component of reducing maternal mortality in low-resource settings. This study aims to identify and understand the determinants of persistently low institutional birth in rural Nepal, with the goal of informing future interventions to reduce high rates of maternal mortality. Methods: Postpartum women giving birth in the catchment area population of a district-level hospital in the Far-Western Development Region of Nepal were invited to complete a cross-sectional survey in 2012 about their recent birth experience. Quantitative and qualitative methods were used to determine the institutional birth rate, social and demographic predictors of institutional birth, and barriers to institutional birth. Results: The institutional birth rate for the hospital’s catchment area population was calculated to be 0.30 (54 home births, 23 facility births). Institutional birth was more likely as age decreased (ORs in the range of 0.20–0.28) and as income increased (ORs in the range of 1.38–1.45). Institutional birth among women who owned land was less likely (OR = 0.82 [0.71, 0.92]). Ninety percent of participants in the institutional birth group identified safety and good care as the most important factors determining location of birth, whereas 60 % of participants in the home birth group reported distance from hospital as a key determinant of location of birth. Qualitative analysis elucidated the importance of social support, financial resources, birth planning, awareness of services, perception of safety, and referral capacity in achieving an institutional birth. Conclusion: Age, income, and land ownership, but not patient preference, were key predictors of institutional birth. Most women believed that birth at the hospital was safer regardless of where they gave birth. Future interventions to increase rates of institutional birth should address structural barriers including differences in socioeconomic status, social support, transportation resources, and birth preparedness. Electronic supplementary material The online version of this article (doi:10.1186/s12884-016-1022-9) contains supplementary material, which is available to authorized users.
