Background and objectives

Accidents are often caused by a complex combination of different factors. A multitude of techniques for accident investigation exist, with their own theoretical underpinning and practical benefits and drawbacks.

This review considers the most widely used accident investigation techniques and identifies their characteristics and usage. It is not a comprehensive review but considers the application, strengths and limitations of the most frequently documented techniques. During the review, the following objectives were pursued:

1. To identify the most frequently used techniques for accident investigation in selected industries, including aviation, nuclear energy, transport, maritime, oil and gas and chemical processing,
2. To understand the characteristics of the techniques in terms of how they conceptualise and represent accident causation,
3. To identify the relevant leading agencies in the six areas and examine what safety accident investigation techniques are being used or recommended by these agencies, and
4. To provide examples of the application of safety accident investigation techniques from previous studies and credible published sources.

Method

The relevant literature was identified and screened in three stages as follows:

Stage I: a systematic desktop search of six major databases was conducted: Google Scholar, Scopus, PubMed, Wiley Online Library, Water Resource Abstracts (ProQuest) and Web of Science.

Stage II: the sources obtained in the first stage were analysed and those with highly relevant contents were shortlisted. The full text of selected sources was subsequently coded and archived for the third stage of examination.

Stage III: The third stage of examination involved checking the selected sources against three selection criteria: (1) application of one or more accident investigation techniques in at least one of the six target industries; (2) whether the source clearly described the application of the original version of investigation techniques; and (3) if it employed the widely accepted investigation techniques. Those sources that did not meet these criteria were excluded from further consideration.

The screened literature was then analysed and classified in four steps as follows:

1. The most widely used or suggested techniques for accident investigation in various high-risk industries were identified.
2. Various characteristics of the techniques identified in step 2 were examined and summarised.
3. Examples of the application of techniques in case studies were identified. In addition, the accident model(s) that underpins these techniques (e.g. sequential, epidemiological and systematic) were identified in this step.
4. Studies that conducted comparative evaluations between different investigation techniques in various high-risk industries were identified and the reasoning (if any) for recommending a preferred technique(s) over others was captured.

Results

The review resulted in the identification of 36 accident investigation techniques across various high-risk industries. The review and comparison of these techniques revealed a diversity of perspectives on accident causation and the benefits and limitations inherent in them. In particular, the academic literature highlights the benefits of a systemic approach to accident analysis, including a more comprehensive understanding of the complexities and dynamics involved in accident causation. Nevertheless, the application of systemic techniques, such as STAMP, FRAM, and AcciMap, has mostly been in the academic domain and practitioners are typically unaware of them or do not use them in their investigations. Some studies have suggested a number of difficulties with the application of systemic techniques in industrial environments, including their theoretical complexity, absence of domain-specific taxonomies to guide the investigation, generally high resource requirements (e.g. time, data collection), high sensitivity of the outcomes to the quality of data, and inconsistency of the results when applied to the same scenario by different users.

Compared to systemic techniques, sequential and linear techniques, such as Fault Tree Analysis and Causal Tree Method, are less resource intensive and easier to apply. Consequently, they have been widely used in practice. However, studies have suggested that these techniques do not effectively capture how human, management and organisational elements combine and contribute to accidents, and can lead to identification of easy-to-find causes and ignore less evident organisational (latent) factors that contribute to accidents.

Conclusion

The advantages and limitations of different accident models and investigation techniques suggest that there may be benefits in integrating or combining features of different techniques to capture different aspects of the accident problem and improve accident investigation outcomes.

A range of different strategies was identified that organisational decision-makers may consider when planning accident investigations. These are:

• Using different techniques during the investigation process can help the investigators to draw on the strengths of different techniques based on the requirements of their organisation and the accident situation.
• Applying multiple techniques can both improve the breadth of the investigation and address the bias that reliance on one perspective can entail.
• The selection of a technique to use, in part, depends on different factors including the aim of investigation, available resources, the investigators’ knowledge and familiarity with different techniques, and an assessment of the complexity and/or severity of an accident. Therefore, organisations may choose to provide “scalable” investigation techniques; including some that are simple and flexible (e.g. Change Analysis, Barrier Analysis, Causal Tree Method), and others that are more in-depth (e.g. Swiss Cheese Model, ICAM, AcciMap).
• The techniques may also need to be adapted, combined and tailored to specific contexts and local needs in order to better meet the specific needs of practitioners in different domains. For example, domain specific knowledge and taxonomies can be integrated into generic systemic techniques such as AcciMap.
• Organisations may establish a process to review investigation quality to ensure that objectives are being met and information that is produced is of high quality, reflects the range of factors involved in an accident and can be used for prevention purposes and organisational learning.
• Organisations need to consider whether the accident investigation technique selected for use allows for the identification of latent factors including system design, management and organisational inadequacies and oversights. The resolution of these distal issues within an organisational environment can help prevent a larger class of accidents than focusing attention on immediate (easily identifiable) causal factors.