Event-based analyses include both accident/near-miss analysis and confidential reporting systems. Each of these two sources of data will be described.
Most organisations have systems for recording safety incidents, accidents and sometimes also near-miss events. These can be analysed to provide data from actual events and can include an examination of the human factors, or non-technical skills, inherent in the accident or near-miss (see Box 9.1). Reports are submitted by employees to the organisation and the more serious accidents will normally have to also be reported to a regulatory authority. Reports of incidents and accidents are usually coded for causal factors by the reporter or by a supervisor or by a specialist in the safety department. In some work settings, the human factors coding of accidents has tended to be minimal. Gordon et al. (2005) argued that many accident reporting systems used by the offshore oil industry in the UK lacked a firm theoretical framework for identifying human factor, or non-technical skill, causes of accidents. In sectors such as health care this recently has begun to change with much more effort devoted to gathering, recording and analysing adverse events (Holden and Karsh, 2007; Johnson, 2006). Accident investigations are conducted for more severe incidents to establish what happened, and to prevent a similar event happening again. although only one diagnostic source, the analysis of accident data, is essential for improving workplace safety (Dismukes et al., 2007; Kayten, 1993; Wiegmann and Shappell, 2003). a post-incident inquiry begins with a negative outcome and considers how and when the defences built into the system failed (hollnagel, 2004).
An accident investigation method based on robust human factors models allows a broader interpretation of accident records, thus potentially reducing the likelihood of future accidents (see Stanton et al. (2005) for tools focusing on human error). Examples of accident analysis tools that include both human and organisational factors are:
- 1. Safety Through organisational Learning (SoL) (Fahlbruch and Wilpert, 1997)
- 2. Human Factors Accident Classification System (HFACS) (Shappell et al., 2007)
- 3. technique for Retrospective analysis of Cognitive errors (TRACEr) (Shorrock and Kirwan, 2002)
- 4. taproot (paradies et al., 1996)
- 5. tripod (Beta and Delta) (Hudson et al., 1994)
- 6. human factors Investigation Tool (HFIT) (Gordon et al., 2005).
Box 9.1 Example of accident analysis in aviation
Accident data were used by orasanu and Fischer (1997) to examine non-technical skills in a study identifying decision-making strategies in the cockpit. Analyses of aircraft accidents were conducted in the uS by the National Transportation Safety Board (NTSB). These analyses (based on crew conversations that are recorded by the cockpit voice recorder (from the ‘black box’), physical evidence, aircraft systems data and interviews with survivors) provide contextual factors that contributed to the pilots’ decision-making, including sources of difficulty, types and sources of error, and decision-making strategies. This information was also combined with data from actual observations in a high-fidelity simulator, along with data from the Aviation Safety Reporting System (ASRS), which is a confidential reporting system maintained by the National Aeronautics and Space Administration (NASA). The use of these three data sources allowed decision situations and decision strategies to be identified, and revealed that unsafe flight conditions were related to failures in pilots’ non-technical skills rather than lack of technical knowledge, flying ability or aircraft malfunction.
Accident analyses are often constrained, as the accident analysis system may be vulnerable to underreporting, have incomplete recordings and may present only part of the overall picture of the event (Hollnagel, 2004; Stoop, 1997). To illustrate, in an examination of US Navy diving accident reports, the largest proportion (70%) of the diving mishaps were attributed to unknown causes; only 23% were attributed to human factors (O’Connor et al., 2007), a rate far below the 80% generally attributed to this source in high-reliability industries.