Quantitative risk assessment studies of accident scenarios usually involve estimating the number of fatalities that can be expected. The number of people injured, however, is seldom evaluated because it implies significant additional effort and often the information required to perform this evaluation is not available. However, the number of injured people can be very important for emergency planning, especially in relatively large accidents. In this paper, a set of 975 accidents were selected for analysis, with the aim of searching for a relationship between the number of people killed and the number of people injured. As the data were scattered, principal component analysis and clustering analysis were applied to identify the data subsets that could undergo a selective, specific statistical treatment. Further treatment of these subsets led to mathematical expressions that are used to estimate the probable number of injured people as a function of the number of fatalities for all accidents, as well as for gas cloud, fire and explosion events, respectively. (c) 2005 Elsevier B.V. All rights reserved.