BACKGROUND: Preterm birth contributes significantly to infant mortality and morbidity, including blindness from retinopathy of prematurity (ROP). Access to intensive neonatal care is expanding in many countries, but care is not always optimal, one factor being that nursing is often by inadequately trained nurse assistants. OBJECTIVE: The aim of this study was to evaluate whether an educational package for nurses improves a range of outcomes including survival rates and severe ROP in 5 neonatal units in Rio de Janeiro, Brazil. METHODS: The study design included an uncontrolled before-and-after study in 5 units, with interrupted time series analysis. Participatory approaches were used to develop a self-administered educational package for control of pain, oxygenation, infection, nutrition, and temperature and to improve supportive care ('POINTS of Care'). Educational materials and DVD clips were developed and training skills of nurse tutors were enhanced. There were two 1-year periods of data collection before and after a 3-month period of self-administration of the education package. RESULTS: Overall, 74% of 401 nurses and nurse assistants were trained. A total of 679 and 563 infants were included in the pre- and post-training periods, respectively. Despite improvement in knowledge and nursing practices, such as the delivery and monitoring of oxygen, there was no change in survival (pre-training 80%, post-training 78.2%), severe ROP (1.6 vs. 2.8%), sepsis (11.3 vs. 12.3 cases per 1,000 infant days) or other outcomes. Outcomes worsened over the pre-intervention period but the change to an improvement after the intervention was not statistically significant. During the study period many trained staff left the units, but few were replaced. CONCLUSIONS: Future studies need to focus on barriers to implementation, team building, leadership and governance, as well as the acquisition of knowledge and skills.