Table 1

The distribution of accident types in the three databases used in this   research。

H。 Malekitabar et al。 / Safety Science 82 (2016) 445–455 449

itational energy that workers gain when they climb a structure, and the more potential energy they gain, the more severe the con- sequences will be (Zalk et al。, 2011)。 Properly installed fall protec- tion systems will certainly ‘‘drive down” the risk of falling, while a steep incline of the working platform drives it up by increasing the probability, and objects with sharp edges on the ground drive it up by exacerbating the consequences。

Despite risk ‘‘sources” that are usually discussed before all else (Tah and Carr, 2000), risk ‘‘drivers” are seldom investigated in tra- ditional risk management approaches (Ashley et al。, 2006)。 There are a few focal points suggested in the literature that may have increased either the possibility or consequences of an incident in construction projects, which are summarized in Table 2。 Literature suggests if there is evidence of such drivers, risks are more likely to arise with higher expected consequences。 Though there is quite lit- tle discussion on how each driver may influence project risks, cat- egorizing all previously suggested drivers will provide a basis for looking for drivers of interest in the special field of safety。

There is no unified framework for thinking about risk drivers。 A certain incident may come  about  through  relatively different paths, each of which comprise different drivers (Graham et al。, 2010b)。 Technical, social or other aspects may be of concern, and external or internal drivers may prevail over others。 To be effec- tively involved in the process of risk identification, a set of key risk drivers  has to  be  established considering:

1。It has to be clarified which risks would be possibly stimu- lated by each driver (Pavaskar et al。,  2012)。

2。Drivers have to be detectable before the risks show them- selves (Berkeley et al。, 1991)。 In the concept of this research, drivers have to be detectable before the design process is complete。

3。Each driver has to be evaluated whether it drives the influ- enced risks up or down (Berkeley et al。,    1991)。

4。Although it is good to make sure that no relevant risk driver is left unattended, a reasonable selection of those reflecting important influencers can prevent unnecessary effort (Bartlett, 2002), and those influencing more than one hazard are  strongly demanded (Orozco Mojica,   2013)。

5。The whole set of risk drivers has to adequately cover all the issues in the project (Berkeley et al。,   1991)。

6。A set of risk drivers has to be an open set。 No one can seal the set, guaranteeing a clear view of all uncertain events (Browning, 1999)。

7。Drivers are to be checked periodically。 Changes in the design, project environment, or knowledge may call for a revised set of drivers。

8。Although no significant superiority can be found among the drivers, it would be useful to know how some drivers may influence the others, e。g。 they might cancel or intensify the effect of each other (Browning, 1999; Graham et al。,   2010b)。

Table 2

Risk drivers found in the literature。 Category Drivers

External drivers ●  Government policies and regulations are constantly changing (Agyakwa-Baah and Chileshe, 2010)

●Urban development has been rapid or unplanned (Cardona et al。, 2012; UNISDR, 2013)

●Project site is located in an underdeveloped region (UNISDR, 2013)

●Theft, vandalism or malicious acts are frequently reported (Agyakwa-Baah and Chileshe, 2010)