H。 Malekitabar et al。 / Safety Science 82 (2016) 445–455 447

found in the analysis with regard to the accident types, and a short discussion of the results and the needs for future research to achieve an efficient safety risk identifying engine forms the last chapter。

2。Material and methods

2。1。Analysis data

A procedure leading to a set of risk drivers has to look at the risks occurring in a project, and analyze form where they arise, what path they follow, and what consequences they have。 This would be more effective if a full diary of events, no matter how trivial they are, were kept in project documentations, because there would be an immediate indication of whether a special event was driving some uncertain hazard up, or should be filtered out。 In the absence of such information, formal analytic reports that inves- tigate construction accidents can be used as a source answering the questions above。

The National Institute of Occupational Safety and Health (NIOSH) has conducted a study on Fatal Accident Circumstances and Epidemiology (FACE) and published more than 300 fatal facts on its website, 248 cases of which are related to the construction industry (the website claims to contain 249 cases but 1 case seems to be duplicated)。 Each case is composed of a vivid narration of the incident accompanied by a discussion on the safety clauses that the constructor is accused of violating。 Another series of investigations with a similar approach is published by OSHA comprising 70 ‘‘fatal facts” that fall in the category of construction accidents。 Both data- bases represent a normal sample that covers all types of accidents and have not been produced during special-purpose programs。

In this paper, the model suggested by Behm (2005) is used to analyze by how much these construction fatalities could have been prevented by a better design。 Behm defined a fatal incident to be related to the concept of DfS, either if ‘‘physical aspects” in the pro- ject were found to contribute to the occurrence of it, or if preven- tive suggestions could have been included in the design process, and reviewing 224 cases presented by NIOSH, he concluded that 42% of the incidents were of such a nature。

Since both databases demonstrate accidents that happened in the United States, an additional database concerning injuries that occurred in a developing country, Iran, was additionally investi- gated。 Forty-nine superintendents and technical experts from var- ious types of construction projects, participating in  a safety training course, were each asked to present one noticeable acci- dent they have witnessed at their work sites。 They were instructed then to analyze the root causes of their case, and find the clauses from both Iran’s safety code and OSHA, violation of which led to the accident。 Forty-five valid cases were collected (four failed to provide the information required for  this  research)。

It is important to note that the accidents here considered did not necessarily claim lives, instead, they mostly ended with multi- ple severe injuries with consequent claims and indirect costs far more than those for a single   death。

To decide whether an accident is related to design, the reports were studied case by case, looking for those events or conditions that may have precipitated each accident as well as any special piece of information about what distinguished the day of accident from every other day of usual work。 Five major things questioned in every accident investigation – who,  what,  when,  where  and how – must be further oriented to special purposes (Reese and Eidson, 2006), and in  this  research,  investigations  were focused on how accidents could be related to    design。

The collected information was refined to exclude three cate- gories from the analysis: (1) aleatory events that happened for