This thesis aimed to understand the factors influencing the reliability of accident and incident classification systems in high-hazard industries and subsequently, to develop methods to achieve high consensus in these classification systems. To explore reliability the Human Factors Analysis and Classification System and its Australian Defence Force derivative version were used as case studies for reliability assessment. The reliability of each system was assessed and factors associated with coder expertise, training and usability explored. The research was then extended across all accident and incident classification systems by using classification theory to guide a strategic review of reliability. This involved qualitative and quantitative analyses of the factors affecting user consensus conducted across more than twenty classification systems across six safety-critical industries. As a result four recommendations were developed for improving the reliability of classification systems. These included designing domain-specific systems, limiting system size and reducing psychological and bias-causing terminology. These recommendations were tested through application to the Australian Defence Force version of the Human Factors Analysis and Classification System. Promisingly, they improved reliability by more than twenty percent across all levels of the classification system. This research has built on previous studies of improving the reliability of accident and incident classification systems that have largely been limited to improvements of specific systems. The recommendations enable a more systematic approach to improving reliability across all accident and incident classification systems within high hazard industries. As a result, the likelihood of effective hazard intervention strategies will be increased and should lead to improvement of safety in high-hazard industries.
Professor Ann Williamson
Dr Nikki Olsen