BACKGROUND Maternal mortality from complications of unsafe abortion constitutes a serious problem in several developing countries. There is, however, a paucity of well-designed and implemented studies in this area, especially in Latin America. The aim of this paper is to present the findings on the determinants and medical characteristics of abortions among women admitted to hospitals. METHODS A descriptive cross-sectional hospital-based study was carried out between October 1992 and September 1993 in Fortaleza, Brazil. A Cox's proportional hazard model was used to estimate prevalence rate ratios after adjustment for confounding. RESULTS Among 2074 (48%) women who admitted to terminating the pregnancy, 66% reported using misoprostol to induce abortion. Women with an induced abortion as compared with those with an unlikely induced abortion are younger, more often not married, have fewer children alive and experienced one or more previous induced abortions. We have not found any important differences with regard to complication or duration of stay in hospital. CONCLUSIONS This finding, at odds with most previous studies, could reflect the special situation in Brazil where misoprostol is used for illegally-induced abortion. The use of misoprostol by this population may have contributed to the reduction of severe complications related to induced abortion which were most prevalent with more invasive methods. Recommendations are made as to the need for confirmatory studies as well as on information regarding cultural perceptions and concepts of abortion, and reasons why poor women fail to adopt available family planning methods.