(47v) Walking the Fine Line Between Fatalities and Injuries

Chris Wells^a  , Alan Dick

^aSuncor Energy, 150 – 6^thAvenue S.W., Calgary, AB, T2P 3E3

PHA Study Teams often have difficulties in assessing the Severity of a Hazardous event or in relating the estimated fatality rate derived from Probit functions to less severe injury rates.  Some guidance is provided in monographs such as the DNV Loss prevention Handbook, which provide general factors relating the number of serious injuries to fatalities or the number of temporary injuries to fatalities.  These factors are generally stated as global ratios and do not shed light on the underlying statistical relationships, nor differentiate any difference based on the type of hazard encountered.  Further information can be gained by analysing the data with more sophisticated methods. 

Recently our team reviewed over 70 industrial accident reports which concurrently reported fatality and injury counts for specific accidents.  These accidents were classified into two categories ( Toxic releases; Fires and Explosions) .  The data was regressed using a cumulative log-normal distribution transformation with good success.  Based on this analysis the team generated a compelling relationship between fatalities and non-fatal injuries.

The sample data shows that both hazard classes are well described by log-normal distributions, but with different statistical characteristics.    The data is in general agreement with some of the less sophisticated ratios available publically, and show that these ratios were probably based on median estimates.  Using this type of analysis allows the analyst to better estimate the median expectation (the most frequent result) ,  or mean expectation ( the average expected result), as well as allowing the estimation of confidence limits around the expectation

During hazard reviews the PHA teams could use these relationships when a fatality has been shown as not credible.