Rating the Severity of Close-Call Events

Abstract

Accidents rarely happen; close calls (“near misses”) are much more prevalent. So learning from close-calls is statistically more feasible. This paper describes a new technique for rating, or having a computer help to rate, the relative severity of close-call events. Various potential applications are in transportation close-calls: between aircraft on the ground or in the air, between highway or rail or maritime vehicles or such vehicles and fixed objects; also close-calls in surgery or other medical treatments on patients, or in police or security work. Currently the technique is being developed and validated for specific application to airport runway incursions – where an aircraft or vehicle violates safe separation criteria relative to another aircraft. The technique starts from severity ranges prejudged by subject matter experts and tabled for each of a set of possible near-miss scenarios and range of closest proximities, including the overall best (global minimum) and worst (global maximum) severities, as well as ratings for the worst case for each of a set of ”influence factors.” (These numbers specify the corners for each dimension of a several-dimensional ”severity space”). A scale of influence levels associated with key objective properties of each influence factor is also prejudged by subject matter experts. Then, for any particular close-call event in any particular scenario, one need only specify the key objective properties of the event. The event's severity is then determined by a combinatorial algorithm drawing on these properties and the tabled values. Validation efforts for runway incursions comparing the technique's predictions to independent expert ratings are promising.