Markov Chain Modelling of Safety Incident Data: A Veritable Decision Support Tool

Abstract

The rate of industrial accident occurrence has remained a perennial
challenge. This paper points to the need for the application of Markov
chain model in unwrapping the deeper meanings buried in the safety
incident data that mere descriptive statistics can hardly furnish. In line
with the study design, a 16-year data was obtained from Nigeria Gas
Company, a subsidiary of Nigeria National Petroleum Corporation
(NNPC). The historical incidence records were characterized and
proved to possess a note of stochastic regularity that fits into a Markov
Chain Model. A twenty state transition was used for the study, namely:
fatality, third party fatality, permanent disability for example. Result
emanating from the study reveals that subjects make about thirteen
habituations among various states in the organization before being
absorbed in any ten absorbing states with a standard deviation of 12.
Remarkably, 70.5% of the field workers in the organization had
noteworthy severe medical treatment case. In conclusion, the Markov
Chain Model was able to identify states such as unsafe acts and unsafe
conditions transitions to have influenced incident levels the most in the
organization. This study has also shown that Markov Chain model can
be successfully applied to industrial accident data, unveiled significant
visits, habituations which the organization can explore in optimizing
their injury prevention programme and ensures field staff safety.