Data from a national survey of pilots was used to examine the validity of measures for the prediction of aviation accidents occurring both prior to the survey and after the survey. Separate retrospective and prospective analyses were conducted, and 45 survey measures were found to be significantly associated with accident involvement in the retrospective analysis. However, only 13 of those 45 measures were found to be significant in the prospective analysis. Most measures found to be significantly related to accident involvement were related to aviation exposure; the remaining measures related to pilots' perceived and actual level of caution. This study is unique in its use of a cohort design for the examination of aircraft accident risk prospectively, and results suggest the need for caution in the interpretation of retrospective analyses in this research domain. Actual or potential applications include the design of aviation safety programs and design or interpretation of studies that address indicators of aircraft accident risk.
Abstract