Which databases for hazard studies?

With the emergence of quantified risk analyses (known as "probabilistic" analyses), databases are proving to be an invaluable and indispensable tool for assessing the probability of occurrence of industrial accidents.

To assess the probability of a phenomenon, the simplest method is to use accident data, based on statistical analysis of events already observed. Another possible approach is to assess probability by calculation, using methods such as the bow-tie, fault tree, event tree or LOPA method. Both approaches require the use of different types of database. This raises the following question: Which databases should be used?