The utility of incident reports is limited by lack of standardized mechanisms for classifying and following up on reports. In this study, investigators developed a classification scheme for surgical incidents based on more than 1000 reports, then developed a structured approach for addressing quality issues raised by the reports. Over a 4-year follow-up period, the frequency of most types of incidents declined significantly. This study, like most studies of data derived from incident reports, is limited because voluntary reports generally are filed only in a small proportion of adverse events. Nonetheless, the study illustrates how to use incident reporting data within an appropriate conceptual framework for improving safety, as discussed in a prior commentary.