Abstracts – Browse Results

Search or browse again.

Click on the titles below to expand the information about each abstract.
Viewing 1 results ...

Alexander, D, Hallowell, M and Gambatese, J (2017) Precursors of Construction Fatalities. I: Iterative Experiment to Test the Predictive Validity of Human Judgment. Journal of Construction Engineering and Management, 143(07).

  • Type: Journal Article
  • Keywords:
  • ISBN/ISSN: 0733-9364
  • URL: https://doi.org/10.1061/(ASCE)CO.1943-7862.0001304
  • Abstract:
    In order to prevent fatalities, new methods of evaluating work conditions and making predictions are needed. Nuclear energy, chemical manufacturing, and commercial airline industries have all used precursor analysis to predict and prevent catastrophic events. This paper presents the first steps toward a precursor-analysis process for the construction industry. First, a comprehensive catalog of 43 potential precursors was established by triangulating results from a literature review; performing a deterministic event analysis of 21 fatalities; and having brainstorming sessions with construction safety, law, regulation, and psychology experts. The 43 potential precursors were then translated into a precursor–data collection protocol. The protocol involved questions and field observations to assess the presence or absence of each precursor before an event occurred. The protocol was applied to collect data for 19 new cases, which included (1) events where high-energy work was successfully completed without incident; (2) near misses where high-energy was released but no one was harmed; and (3) fatal or disabling injury events. Using these cases, a controlled experiment was conducted where a group of 14 experts were asked to predict each case outcome using only the leading information collected via the protocol and their judgment. Later, the same experiment was conducted with moderately experienced professionals and students for validation and to test generalizability. A permutation test indicated that people of all levels are able to distinguish between success and failure far better than random using only the leading information. Future research is proposed to reduce the scope of the protocol and to create objective methods of prediction using statistical tools, thereby making the precursor-analysis process less resource intensive and more reliable.