Background Systematic analysis of error recovery can provide hospitals with important information to help them improve their ability to detect and correct errors. Because errors will always crop up and 100% safety can never be achieved, hospitals should be able to prevent patient harm by timely and effective error recovery. Methods In this study, failed, missed and absent recovery opportunities were identified in 52 medication errors which all resulted in severe patient harm or patient death. For all identified recovery opportunities, the underlying failure factors were identified and classified according to the Eindhoven classification model. Those failure factors represent negative influences on error recovery. Results The number of recovery opportunities per error ranged from 0 to 11; on average, 2.4 recovery opportunities were identified. Of 127 identified recovery opportunities, 94 (74%) were planned and 33 (26%) were unplanned or ad hoc. Most failure factors underlying the planned recovery opportunities were organisational failure factors; most failure factors underlying the unplanned recovery opportunities were human failure factors. Conclusions From this study, it can be concluded that actual accidents can be used as an alternative data source to near misses for the analysis and understanding of error recovery. By using both sources, hospitals can enhance their resilience by reinforcing the positive influences on error recovery as well as reducing the negative ones. Together with traditional error reduction methodologies, which only concentrate on eliminating failure factors, hospitals thus have numerous opportunities to improve patient safety.