thèse doctorat de l'université bordeaux 2 - ISPED-Enseignement à ...

variables included socio-demographic variables (age, sex, marital status, education, andemployment), whether driving permit was issued without practical test, frequency of reportedrisky driving behaviors (sleepy driving, cell phone use while driving, seat-belt use, traffictickets, driving while intoxicated during previous three months), and involvement in RTCduring previous year.Using 17-inch video screens, five test videos (three from Pakistan and two fromCameroon) were shown to drivers before presenting selected sites. The order of sites wasrandomly drawn for each participant. To avoid confusion from right- and left-hand drivingpracticed in Cameroon and Pakistan, site videos from Cameroon followed those fromPakistan. For each video shown, drivers were asked to report their perception of site andtraffic, on a four-level scale; 1/ Certainly safe, 2/ Probably safe, 3/ Probably dangerous,4/ Certainly dangerous. Further, they were asked to record their preferred speed (in km/h) foreach site.Each site was characterized by the main investigator, using definitions used in ourprevious study conducted in Cameroon (Bhatti et al.). Site-related variables assessed werebuilt-up or rural area, horizontal and vertical road profile, road width, surface regularity, vergeslope, depth at 10 m from the verge, location and type of nearby obstacles (within a roaddistance of 50 m in each direction), horizontal marking, vertical road signs, and presence ofan intersection or a U-turn. Traffic-related variables assessed were traffic moving in same oropposite direction, visible pedestrian, motorcyclist, or heavy vehicle, rain or wet surface,maneuvering vehicle (crossing or overtaking), and number of lanes (Sümer, Ünal, Birdal,Çinar, & Çevikoglu, 2007).AnalysisProportions of site- and driver-related characteristics were computed. Discordance (D)of appreciation for a matched high- and low-risk site pair was defined as “minor” whendifference of hazard perception level was one on the Likert scale and “major” when the leveldifference was more than one. Positive sign (D + ) was used to show that hazard perceptionlevel was higher for the high-risk site than its matched low-risk site, and negative sign (D - ) toshow that hazard perception level was lower for high-risk site than its matched low-risk site.Wilcoxon test was used to assess whether these discordances were significantly higher orlower for high-risk site than low-risk site. Similarly, differences in reported speeds formatched high- and low-risk site pairs were compared using a paired t test. Correlationsbetween reported speeds for high- and low-risk site pairs were assessed by intra-classcorrelation coefficient (ICC).Associations of driver-, site-, and traffic-factors with road hazard perception levelwere assessed using logistic regression (model 1) with a backward selection strategyincluding significant (P