Voluntary error reporting systems are ubiquitous in health care, but such incident reports often fail to result in lasting safety improvements. Prior studies have shown that most hospitals do not have robust mechanisms for analyzing and learning from reported errors, and voluntarily reported errors may represent only a narrow part of all errors. In this study, the authors developed a statistical model to analyze incident reporting data to identify work areas with safety problems and rank the severity of safety problems identified. The model was validated using data from the National Reporting and Learning System. This model, once further validated, could greatly aid large organizations in using incident reporting data to improve safety.