| A simple method for a contractor to predict the impact of construction accidents does not currently exist. This research looks at the efficacy of statistical modeling of construction safety data, investigates methods to model construction safety performance, and presents tools to predict the impacts of accident losses in compensation weeks. This is done for a given source and nature of injury or illness, accident type, disability type, occupation, part of body involved, and age of injured. The results of this research have demonstrated the efficacy of a few modeling approaches, exposing the potential for continued research and refinement to suit the user industry. The New York State Workers' Compensation data of 91,953 cases of construction injuries and illnesses closed between 1980 to 1988 serve as the basis for this research. Linear regression modeling is investigated and suggestions for improvement of such a model are presented. The introduction of classification and regression trees (CART) as a tool for simplified analyses of the impact of construction-related accidents is presented. The concepts for the models developed herein can be used to predict the impact of construction work site accident losses in compensation weeks. Refined versions of these regression and CART models may prove useful to many in the construction industry including insurance providers, contractors, safety professionals, construction engineers, construction managers, and design engineers. |
