A Method to Calculate the Accident Probabilities in Construction Industry Using a Poisson Distribution Model

Occupational accidents not only lead to serious damages or loss of life and money, but also cause a serious reduction of the productivity. Existing occupational health and safety practices are not sufficient to prevent the construction accidents in Turkey. Especially, in construction industry the accident rate is five times higher than the employment rate. Benefiting from more realistic and objective risk assessment methods can help to reduce the accident rates. Therefore, accident probabilities should be calculated objectively without considering personal experiences or subjective assessments. In this study an objective and quantitative accident probability calculation approach is proposed. Firstly, the accident probability is redefined. Based on this definition the accident probability became a function of accident rates and worker exposure values. The accident rates were calculated from statistics of real accidents. In order to determine accident rates expert witness reports that were examined for 13 years (2000-2013) and the statistics that were presented by Social Security Institution were used. In total 623 expert witness reports that belonging the actual construction work accidents between 2000 and 2013 were analyzed. The exposure values were calculated from the man-hour values that were taken from project schedules and planning tables. Then, accident probabilities were calculated by using a Poisson distribution model where accident rates and exposure values were used as a distribution parameter. Within the scope of this study, instead of focusing all construction activities it was decided to focus on 5 main activities which show the highest accident rate in order to provide better results to prevent accidents. These five main activities were chosen as; Excavation, Reinforced Concrete Works, Masonry, Plaster and Painting Works, Roof Works. In an industry like construction industry where the accidents are recorded in non-specific standards, it is very difficult to represent probability of accidents with a known distribution model. Thanks to the approach presented in this study it was possible to provide an objective method to obtain accident probabilities using Poisson distribution.