Kathryn Woodcock

How can we evaluate, improve, and support accident investigation?

Accident investigation is a prominent component of health and safety programs and a highly valued method of understanding where hazards exist. This research area is developing methods to evaluate and support investigation particularly for cases involving human error, across a variety of domains.

This research area is developing methods to evaluate and support investigation particularly for cases involving human error, across a variety of domains.

• A modified Haddon Matrix notation was developed to evaluate investigations.
• A simulated investigation technique is used to learn, practice, and evaluate investigations.
• A tool for human error data acquisition has been developed and is being tested and refined to increase understanding of specific human error patterns involved
• A data collection tool has been developed to expand information search in investigation.

This work has been carried out with various forms of support and sponsorship including Hydro One Networks Inc. and the U.S. Federal Aviation Administration human factors program in addition to the amusement industry.

Content analysis of 100 consecutive media reports of amusement ride accidents

Woodcock, K., 2008. Accident Analysis and Prevention 40, 89-96.

Accident investigations influence public perceptions and safety management strategies by determining the amount and type of information learned about the accident. To examine the factors considered in investigations, this study used a content analysis of 100 consecutive media reports of amusement ride accidents from an online media archive. Fatalities were overrepresented in the media dataset compared with U.S. national estimates. For analysis of reports, a modified "Haddon matrix" was developed using human-factors categories. This approach was useful to show differences between the proportions and types of factors considered in the different accident stages and between employee and rider accidents. Employee injury accounts primarily referred to the employee's task and to the employee. Rider injury reports were primarily related to the ride device itself and rarely referred to the rider's "task", social influences, or the rider's own actions, and only some reference to their characteristics. Qualitatively, it was evident that more human factors analysis is required to augment scant pre-failure information about the task, social environment, and the person, to make that information available for prevention of amusement ride accidents. By design, this study reflected information reported by the media. Future work will use the same techniques with official reports.

Using simulated investigations for accident investigation studies.

Woodcock, K., Drury C.G., Smiley A.M., Ma, J., 2005. Applied Ergonomics. 36: 1-12.

Regardless of the actual causes of particular accidents, it is the causes identified by the analyst that determine what responses are made, and how safety is managed in industry. Past authors have suggested that investigation might be biased, but studies were limited by the lack of similarity to real-world investigation tasks in which investigators must decide what information to acquire as well as analyse and interpret it. A technique was developed to use simulated investigations rather than attribution judgements about causation. Three studies are described, using simulated investigation to compare elicited knowledge and hypotheses among safety specialists, to compare investigations using job aids with unaided investigations, and to teach students about investigation bias and comprehensiveness. The method was well accepted by participants and shows flexibility for a range of uses, although it may have limitations.

Comparison of simulated accident investigations performed by groups comprised of homogenous "type"

K. Woodcock, J. Tsao, 2005. Proceedings of the Association of Canadian Ergonomists. [CD-ROM] 5 pgs.

A group of 300 safety representatives from the same organization were divided into groups with a homogeneous personality type for a simulated investigation of a hypothetical accident in their industry, following a training seminar on the types and mechanisms of human error. Consistent with past studies, the "SJ" temperament was most common. Within-type similarities and between-group differences were not as apparent quantitatively as qualitatively. Most groups retrieved about half of the story facts, regardless of causal relevance. However they cited few of those facts in their conclusions, which were dominated by inferences about "root causes". All groups tended to avoid ascribing error to the injured person, despite inclusion in the story of that person's non-culpable errors. The results suggest that randomly assembled investigation groups are not assured a diversity of perspectives, and that all types will need further aiding to extract more information about human error from accident investigations.

Toward a tool for human error data acquisition (THEDA) for investigators

K. Woodcock, 2007. Proceedings of the Association of Canadian Ergonomists. [CD-ROM] 6 pgs.

There are many models of accident causation and human error that can explain and classify event data. However investigations in specific cases often fail to acquire all of the information needed to apply the models to an event. THEDA was developed as an interactive tool to aid the investigator to seek additional data about human errors involved in accidents and hazardous device failure situations. Using a hierarchically branching adaptive sequence of yes-no questions, THEDA promotes acquisition of data by prompting the investigation to probe into procedures, systems, conditions, and actions that created the situations in which the proximal individual responded ineffectively. A functioning mock-up has been created in spreadsheet form which compiles the user's findings for reporting with general human factors comments. Although operational for trained users, its development is ongoing to enhance its general usability.

Comparison of common methods for accident data classification: case study in a Canadian utility company.

J. Tsao, K. Woodcock, 2007. Association of Canadian Ergonomists. [CD-ROM] 6 pgs.

Accident data are commonly used to review company safety performance and plan countermeasures. The perceived safety performance and problem areas are influenced by the method of extracting, collecting, and summarizing data obtained in the accidents. A Canadian utility company was reviewing options for new or additional methods of accident data recording. The methods were compared qualitatively using historical data from the company. Each accident report was broken down into single fact data points and recorded as factors without assuming any information not stated. This analysis suggested that the classic Haddon matrix is worth further consideration.

Measuring the Effectiveness of Error Investigation and Human Factors Training

C.G. Drury, J. Ma, K. Woodcock, 2002. FAA HF Workbench

This report provides the findings from the final year of a three-year study of how effectively aviation maintenance errors and incidents can be investigated. It is important for aviation safety that errors, incidents and accidents be investigated thoroughly to learn the correct lessons to prevent future incidents. While much necessary effort has been focused on analysis of the causes of errors, these analyses ultimately depend for their validity on whether or not the appropriate set of facts was collected by the investigators. During prior years of this project it was established that investigators only collect a fraction of the available facts. The current project was designed to measure the effectiveness of job aids in improving the thoroughness of investigations.

A New Model of How People Investigate Incidents

Drury, C. G.; Woodcock, K.; Richards, I.; Sarac, A.; Shyhalla, K., 2001. SUNY at Buffalo.

We used a simulation methodology to provide a direct measurement of how incidents and accidents are investigated. Thirty-seven aviation maintenance personnel with incident investigation experience investigated are six incident scenarios that we developed from actual maintenance incidents. Using a methodology developed by Woodcock and Smiley (1999), participants were given a brief incident description and had to question the experimenter to determine how the incident happened. We counted the number and types of information requests, and recorded their sequence. Based on the sequence data we propose a five-stage model of incident investigation. An initial trigger initiates an interactive data collection/ data analysis period, starting by determining spatial and temporal boundaries then investigating in a somewhat sequential manner. A stopping rule is used to determine when to stop investigating and begin the final reporting stage. The number of facts considered grows during the investigation stage, but then decreases at the reporting stage. Thus, basing recommendations on the reports of incidents may not consider all causal factors.

Do Job Aids Help Incident Investigation?

Drury, C. G.; Ma, Jiao; Woodcock, K. 2001. SUNY at Buffalo.

A previous study established that investigators only collect a fraction of the available facts, and further select facts for their reports. The current study was designed to measure the effectiveness of job aids in improving the thoroughness of investigations of incidents in aviation maintenance. The methodology involved having participants investigate a known incident scenario by asking the experimenter for facts, as they would in their normal investigation routine. The two job aids used were the Maintenance Error Decision Aid (MEDA) developed by Boeing and the Five Principles of Causation (Marx and Watson, 2001). Both are used extensively in aviation maintenance. We tested a total of 15 experienced users of the two job aids, where the investigators were provided with the job aid they had been trained to use. Eleven of the 15 participants used their job aids during the investigation but four did not. The results showed a significant improvement in investigation performance when the job aids were actually used.