Using a Data-Driven Method of Accident Analysis: A Case Study of the Human Performance Reliability (HPR) Process

Abstract

Human error and its contribution to occupational accidents and incidents has received considerable research attention in recent years. However, more research is needed into the validity, practicality, and functionality of using data-driven accident/incident analysis methods to identify factors that contribute to incidents with the greatest frequency. This paper presents a case-study of one such method: Human Performance Reliability (HPR). Methods: The authors conducted approximately 30 HPR reviews to analyze incidents that occurred at a large refining company over a three year period. Through the HPR process, the authors identified the most common human errors, other contributing factors, and the controls (SOPs, processes, programs) that failed to prevent the accidents/incidents. Results: A Chi-Square Goodness-of-Fit test and post-hoc analysis of Standard Residuals on the human error frequencies revealed the most common human errors and contributing factors, while raw frequency counts showed the most commonly associated controls (see Tables 3-6). The Chi-Square statistic was X2 = 528.58, indicating that certain errors were contributing to incidents significantly more often than others. Discussion: Early evidence supports the notion that the HPR process is an effective tool for incident analysis and subsequent continuous improvement efforts in process safety.