Extraction and analysis of construction safety hazard factors from open data

Abstract

Construction is one of the most injury prone industries worldwide. Concerns of health and safety of the employees in construction sites have been a vastly discussed topic for decades. In many countries, companies are required to report safety incidents by using catastrophe investigating report in their workplaces to relevant authorities whereby such data is made publicly available under the open data policy. These open datasets may be well structured or may require further preparation in order to be usable. Some datasets are in the form of reports, which require qualitative, textual analysis to extract insightful information. The purpose of this study is to extract safety hazard factors from an open dataset obtained from the US Occupational Health and Safety Administration, and to further analyse such factors using statistical analysis techniques. For each reported case, text analysis was carried out with the narrative data field describing the circumstances leading to safety incidents to extract safety hazard factors. These hazard factors were categorized into human factors, technical factors, external environmental factors, organizational factors and other factors. The results showed that hazards related to human factors are most common. Descriptive statistics also showed that the most frequent nature of accident was fractures and most frequently occurring accident event was falls to the lower levels. Such information can help to provide insights into the accidents occurred and how relevant authorities may devise strategies to improve construction site safety.