The analysis of UK railway accidents and incidents : a comparison of their causal patterns

• Main Author: Wright, Linda B.
• Published: University of Strathclyde 2002
• Subjects: 388; Urban planning & rural planning
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.269869

An essential assumption for the usefulness of near miss reporting is the common cause hypothesis: the causal pathways of near misses are supposed to be similar to those of actual accidents (such as injuries and damages). The common cause hypothesis was originally proposed by Heinrich (1931) in his seminal book "Industrial Accident Prevention". Since then, the hypothesis has been alternately supported and rejected based on a confounded view of the interdependence of severity, frequency and causation. The evidence from various studies is examined and it is concluded that the hypothesis has not been properly tested. Thus this thesis tests the validity of the common cause hypothesis. In order to develop the methodology to test the ommon cause hypothesis analytical work in the area of incident analysis and reporting was required. Thus this thesis also outlines the approaches to accident and incident analysis and makes several recommendations regarding the use of taxonomies and reporting systems. A reporting and analysis system (CIRAS) for the collection and analysis of near misses and unsafe acts and practices was developed and implemented for use in the UK railway industry. This reporting and analysis system formed the basis for the test of the common cause hypothesis. Data used to empirically test the common cause hypothesis come from one company of the UK railway industry. Three types of data were used: incidents resulting in 'fatality & injury', 'damage' or 'near miss''. A total of 240 incidents were collected via management reports and a voluntary reporting system. All incidents were coded for causal factors according to the CIRAS (Confidential Incident Reporting and Analysis System) taxonomy. A total of 750 causal factors were assigned to the 240 incidents. Analysis was performed on a comparison of the proportion of codes occurring at all three consequence levels using Chi-square analysis. Results : The CIRAS taxonomy consists of 21 individual causal factors. Only three of these factors (knowledge based, training and procedures) were significantly different across the three severity levels. It is therefore concluded that this research provides qualified support for the common cause hypothesis.