Measuring Changes in the Distribution of Incident-Outcome Severities: A Tool for Safety Management
AbstractResponsible organizations monitor events that do or could injure their workers, or members of the public or the environment. As now is widely recognized, not all hazards pose the same relative risks that, if there is an incident, the outcomes will be serious or severe. The literature is found to not offer a clear method for assessing, statistically, upward (or downward) shifts in the distributions of incident-outcome severities that arise under different circumstances. An appropriate method could assist analysts who compare risks in diverse contexts—whether from floods, industrial hazards, hurricanes, or even financial risks. Typically, adverse-incident data are dichotomized (e.g. into fatal versus non-fatal outcomes), but advantages will be shown for a method that allows entire distributions of severity-outcomes to be compared. Applied to case-based examples from industry and a government safety organization, two methods are presented (at an Intermediate level) for comparing severity-outcome risks—one based on resampling procedures and the other using parametric approximations. Besides the data, the Excel file for this paper includes automated templates for readers to apply and experiment with the proposed methods.