Learning from Accidents: A Systematic Review of Accident Analysis Methods and Models

Learning from Accidents: A Systematic Review of Accident Analysis Methods and Models

Hans Wienen (University of Twente, Enschede, Netherlands), Faiza Allah Bukhsh (University of Twente, Enschede, Netherlands), Eelco Vriezekolk (Agentschap Telecom, Groningen, Netherlands) and Roel J. Wieringa (University of Twente, Enschede, Netherlands)
DOI: 10.4018/IJISCRAM.2018070103
OnDemand PDF Download:
No Current Special Offers


After a risk has manifested itself and has led to an accident, valuable lessons can be learned to reduce the risk of a similar accident occurring again. This calls for accident analysis methods. In the past 20 years, a large number of accident analysis methods have been proposed and it is difficult to find the right method to apply in a specific circumstance. The authors conducted a review of the state of the art of accident analysis methods and models across domains. They classify the models using the well-known categorization into sequential, epidemiological, and systemic methods. The authors find that these classes have their own characteristics in terms of speed of application versus pay-off. For optimum risk reduction, methods that take organizational issues into account can add valuable information to the risk management process in an organization.
Article Preview

1. Introduction

The research reported in this article was motivated by the goal of defining a risk analysis method for telecommunication service failure. Failure of telecommunication services can both cause crises (e.g. loss of signaling along railroads) and aggravate crises due to the inability to effectively communicate during the crisis resolution stage. In previous research, we have defined the Raster method for risk analysis of telecommunication service failure, which is now used by the Dutch telecommunications regulator (Vriezekolk et al., 2011), (Vriezekolk et al., 2015). Our current research aims to strengthen risk analysis by including lessons learned from past accidents. As part of the Linc (learning from incidents) project started by the Dutch telecom regulator, we conducted a comprehensive literature review of accident analysis methods, which we report here.

Methods to investigate incidents and accidents have been proposed and investigated at least since the early 1970s, but the number of proposals has increased steeply since the late 1990s (see Figure 4). Different methods have been proposed for different domains (such as DREAM - specifically for road traffic, and ATSB, for railroad infrastructure), and they may use different incident and accident models. For quite a few methods, there is little published evidence that they have been used more than once or twice. A few methods are cited frequently (e.g. STAMP, AcciMap, see Figure 2), and some techniques, such as fault tree analysis, are widely used in practice. Some methods are easy to use but lead to a potentially incomplete analysis, while others are expensive to use while allowing a more thorough analysis. This raises the following questions: (i) what is the state of the art in accident analysis, (ii) can we distill a generic method from all of the published methods, using the useful elements of all published methods, and (iii) can we provide guidelines for choosing an appropriate accident model depending on the context of the incident or accident. In this paper, we answer these questions.

We present a systematic literature review of incident and accident analysis methods published since the early 1970s across many domains. We start with a review of incident and accident terminology in different domains in section 2. We then describe our research method, including our research questions, our systematic review approach, strategy and execution in section 3. In the Results section (section 4), we introduce the three classes in which these methods are generally categorized, and we describe some statistics of the corpus of literature we found. We provide qualitative analytics for the different classes of analysis models. Finally, in the discussion and conclusion section (section 5) we formulate the answers to our three research questions. We present a generic incident and accident analysis method, that can be applied in a lightweight to heavyweight manner, depending on the context of the incident or accident. We also provide generic incident and accident models for two of the major classes of analysis methods: sequential and epidemiological. A full account of our literature review, including selection method, databases under investigation, the full corpus under consideration and an overview of all methods found has been reported in a technical report (Wienen et al., 2017).

Complete Article List

Search this Journal:
Open Access Articles
Volume 11: 2 Issues (2019)
Volume 10: 4 Issues (2018)
Volume 9: 4 Issues (2017)
Volume 8: 4 Issues (2016)
Volume 7: 4 Issues (2015)
Volume 6: 4 Issues (2014)
Volume 5: 4 Issues (2013)
Volume 4: 4 Issues (2012)
Volume 3: 4 Issues (2011)
Volume 2: 4 Issues (2010)
Volume 1: 4 Issues (2009)
View Complete Journal Contents Listing