Naturalistic Studies (Driving In Traffic)
Ramaekers (2003) reviewed 10 studies conducted to determine the effects of anti-depressants on actual driving performance, using a standard road test. In this standard test, a subject drives over a 100-km (62 mi) circuit on a primary highway in normal traffic. A licensed driving instructor rides in the front passenger seat in a vehicle with dual controls. During the one-hour drive test, the subject is instructed to maintain a constant speed of 95 km/h (59 mph) and a steady lateral position within the boundaries of the slower-traffic lane.41
The primary performance measure in this road test is an index of road tracking precision or “weaving”: the standard deviation of lateral position (SDLP). SDLP is measured in centimeters, using an electro-optical device mounted on the rear of the vehicle which continuously records lateral position relative to lane-line delineation. A justification for using this measure is that in studies of the effects of marijuana on driving, SDLP has been found to be a more sensitive indicator of impairment than measures of following distance and general driving proficiency (Ramaekers, Berghaus, van Laar, and Drummer (2004). Ramaekers (2003) further states that SDLP is a very reliable characteristic of an individual’s normal driving behavior, with high test-retest reliability, and is very sensitive to sedative drug effects. As such, it at least deserves consideration as a fitness-to-drive outcome measure for a wider range of drug studies.
It is important to note that SDLP has not been validated against actual crash involvement. Instead, performance decrements on this measure have been calibrated to blood alcohol concentration (BAC), which in turn is highly correlated with crash risk. An alcohol calibration curve was established, showing that drinkers’ mean SDLP rose exponentially with BAC. Epidemiologic studies have shown that a BAC greater than 0.50 mg/ml is associated with an exponential rise in the relative risk of fatal traffic crashes (Borkenstein et al., 1964).42 Ramaekers (2003) asserts that the results from the alcohol calibration study can be used to describe drug effects on SDLP in terms of BAC equivalents. For the standard driving test—which has not changed substantially in 70 studies spanning two decades—the change in SDLP associated with a BAC of 0.5 mg/ml, a 2.4 cm deviation in lateral lane position, is taken as the lowest criterion value defining drug-induced driving impairment.
Interestingly, Ramaekers (2003) found only relatively modest correlations between subjects’ performance on seven conventional laboratory tests of psychomotor performance (e.g., divided attention, perceptual motor coordination, sustained attention, and working memory), subjective ratings of drowsiness, and performance on the standard driving test. The highest intra-subject correlations were found between SDLP and tracking test performance in the laboratory.
The research reported by Ramaekers (2003) is noteworthy with respect to its use of an objective measure of continuous behavior while driving. Also of interest are on-road evaluations that score behavior in categorical terms (e.g., checked mirror versus did not check mirror before lane change); such studies in this review include Roenker et al. (2003), Janke (2001), Janke and Hersch (1997), DiStefano and MacDonald (2003), Mallon and Wood (2004), and Freund et al. (2002). Finally, many road test protocols rely exclusively on examiner ratings or other subjective means of scoring driver performance; studies of this nature that are reviewed below include Richardson and Marottoli (2003), Wild and Cotrell (2003), Duchek et al. (2003), DeRaedt and Ponjaert-Kristoffersen (2001), and Mallon and Wood (2004).
Roenker et al. (2003) designed a driving evaluation based upon a review of crash literature, traditional drivers’ tests, and fitness-to-drive evaluations. They employed an on-road course, consisting of two loops through a 7-mile urban/suburban route. Repetition of the route was conducted to provide an opportunity for a range of traffic conditions to occur. The route permitted observation of maneuvers identified in the literature as especially difficult for older drivers (e.g., left turns across traffic). For each location where a potentially difficult maneuver was attempted, two back-seat raters coded the extent to which the driver’s behavior constituted a dangerous maneuver—one which either required the driving instructor to take control of the car or one where other vehicles had to alter their course in order to avoid a collision. Three raters were trained on a total of 455 items until all scoring criteria could be consistently applied by all raters (interrater reliability was r ≥ .92).
The following driving behaviors were rated on a scale of 0 (very unsafe) to 2 (safe or appropriate) in this research: maintaining lane position, activating signals, stopping smoothly, searching (additional mirrors were placed in the car to assist the raters in detecting eye movement), selecting gaps, accelerating and decelerating smoothly, turning, maintaining speed, maintaining position in traffic, and dangerous maneuvers. Stopping at a stop sign was rated on a four-point scale, where running the stop sign = -1, rolling through a stop sign = 0, stopping at an inappropriate position = 1, and 2 = an appropriate stop. After the drive, the raters also provided a global rating of the drive ranging from 1 (drive aborted/very unsafe) to 6 (very competent/safe).
The Roenker et al. (2003) driving evaluations were performed in a car modified with a passenger-side brake pedal for use as needed by a driving instructor. The driver was familiarized with the car prior to beginning the evaluation. Drivers were informed that they should drive as they normally would, and would follow route instructions given by the driving instructor. The driving instructor followed a script to ensure that instructions were standardized across drivers. Drivers were directed through a 1-mile warm-up route, and then drove through both loops of the 7-mile course. The evaluation was conducted during daylight hours and required 50 to 60 minutes to complete. Subjects consisted of 95 licensed drivers ranging in age from 55 to 86 with a mean age of 71.
The driver performance data were reduced by grouping a total of 455 items into 13 composites, based on the behaviors being rated. Four composites—gap selection, acceleration, deceleration, and right-of-way—were subsequently dropped from the analysis due to ceiling effects (performance was nearly perfect on most observation opportunities). In addition, the composite measure for “search” was dropped for lack of sufficient data, reflecting difficulty in reliably scoring this behavior. The dropped measures are shaded in the list below. The remaining eight composites, italicized below, were scored according to the multilevel rating system described earlier, except for dangerous maneuvers, which were scored as an actual number of observed events.
41 A 1995 “Note for Guidance” on psychotropic drugs in the European Union stressed that tests for assessing driver fitness should minimally last 1 hour, because motivational factors may affect the results of the tests.
42 To convert to BAC terminology used in the United States, divide by 10; thus, 0.50 mg/ml = .05 grams per deciliter. [Consider just plugging in the .05 BAC number and omit the mg/ml conversion problem.