Table of Content
The authors wish to acknowledge the crucial assistance of the California Department of Motor Vehicles in the performance of this study. Most critical were the contributions of Dr. Mary K. Janke, who was responsible for development of the drive test protocols and who coordinated analysis of an important subset of data included herein; and Ms. Sandra Winter Hersch, a graduate assistant employed by Dr. Janke, who administered the functional test battery. Continuing support from Mr. Raymond C. Peck and Ms. Carole Bedwell throughout the duration of this project are also gratefully acknowledged.
Table of Content
The goals in this research were to obtain valid field measures of older drivers' difficulties when negotiating intersections, and to determine if their visual, mental, or physical abilities measured in an office could predict their performance behind the wheel. This work was spurred by statistics showing intersections to be the singular site type where older drivers are most significantly overrepresented in motor vehicle crashes.
Field observations of intersection negotiation were conducted using 82 subjects referred to the California Department of Motor Vehicles (DMV) for special testing. Preceding the field study, a literature review produced a synthesis of age and intersection driving difficulties, which is available as a companion volume to this report. This review guided development of a functional test battery, termed MultiCAD. A task analysis of intersection approach and negotiation requirements for safe vehicle control actions was also completed.
Of the 82 subjects, 26 were labeled as cognitively impaired by the referral source, and 56 entered the study labeled as cognitively unimpaired. The average age of all subjects was 77. Each subject completed a battery of functional measures to test vision, attention, and selected perceptual skills. Specifically, the functional abilities of the study sample measured by the test battery (MultiCAD) included static and dynamic visual acuity; static and dynamic visual contrast sensitivity; sensitivity to relative motion of other vehicles slowing or stopped in the road ahead; divided attention (in a brake reaction situation); detection of pedestrian and vehicle targets in the visual periphery, while attending to a central (foveal) task; and head/neck flexibility (degrees of rotation to both sides).
Following completion of the functional test battery, all subjects performed test drives over a common "standard" route of relatively low familiarity. Unless terminated for safety reasons, the subject then completed a test drive over a high familiarity route in his/her home area. On both routes, the subjects used their own vehicles, and were accompanied by a DMV examiner. The examiner could terminate a test drive at any time for safety reasons.
Driving errors were recorded both by in-vehicle cameras installed on the subject's vehicle by the project team, and by the DMV examiner using a score sheet. Key findings are summarized below.
The video-based classification, which expresses how often an error occurred in relation to the total number of opportunities to commit the error, showed that scanning errors were extremely common. Drivers failed to observe behind their vehicles before slowing down during the approach to an intersection 87 percent of the time on unfamiliar routes and 96 percent of the time on familiar routes. They also failed to scan to the sides after entering the intersection 75 percent of the time, on both route types. One type of maneuvering error, "infringing on others' right-of-way when changing lanes," was also notable, occurring at a 90 percent rate on unfamiliar routes and a 57 percent rate on familiar routes. This error was operationally defined as initiating a lane change with less than 2 seconds of headway.
Further analysis showed mean (videotaped) error rates to vary as a function of the type of traffic control (signal, stop sign, yield, or no control), the familiarity of the course, and the type of movement (straight through, left turn, right turn). Route familiarity had little to no effect on error rates exhibited at signalized intersections. However, for right turns in yield and uncontrolled intersections, error rates were noticeably higher on the unfamiliar course. This may have resulted from drivers "knowing what to look for" as a result of experience in familiar areas.
The error classification produced from the DMV examiners' score sheets followed the same general pattern as the video-based error classification, where scanning errors predominated across both familiar and unfamiliar test routes, and maneuver errors occurred less frequently. The more common maneuver problems included "failure to come to a complete stop at a stop sign," which was noted on 53 percent of the test drives over unfamiliar routes and 57 percent of the test drives over familiar routes. "Stopping for no reason" was noted on 39 percent and "turning too wide or too short" on 46 percent of test drives over unfamiliar routes; both were noted on 26 percent of test drives on familiar routes. Other potentially serious safety problems noted on at least 20 percent of test drives (on unfamiliar routes) by the examiners included "stopped over limit lines (stop bars)," "consistently drives too slowly," and "unsafe left turn gap acceptance." These errors were less common on the familiar routes; presumably this reflected differences in drivers' expectancies for the demands encountered along each route type. Errors that were more frequent on the familiar routes, being noted on at least 20 percent of test drives, included "infringes on others' right-of-way when changing lanes" and "near miss (pedestrian or car) other than during gap acceptance." It is important to note that errors recorded on examiners' score sheets were not exposure-based. Whether an error was committed once or many times by a driver, it was recorded only as a single occurrence, and the number of opportunities to commit the error are not reflected in these data.
To determine the efficacy of the MultiCAD tests in predicting on-road driving performance, correlational analyses were performed to determine the strength of the relationship between each functional status test in the MultiCAD battery and an error score derived from the DMV examiner's score sheet. The error score was weighted to reflect the seriousness of the committed errors. Results of the analyses showed that the weighted error score was significantly predicted by (1) the speed of (correct) responses on certain measures of visual acuity and contrast sensitivity, and (2) response accuracy (error rate) on brake reactions to pedestrian and vehicle movements constituting safety threats that were presented in a driving video.
This report discusses implications of the present research findings for licensing assessment strategies and policies. Older drivers, like all drivers, seem to engage in many intersection negotiation behaviors that could be classified as driving errors, but which have little apparent bearing on safety. Therefore, research into the types of predictor-criterion relationships at issue here should focus specifically and exclusively on those errors which best predict crashes, consistent with the practices of licensing examiners. Reports of significant relationships with performance measures that do not discriminate crashes only cloud the issue.
The present findings suggest that improvements in the safety of intersection negotiation by older drivers can be brought about through changes in engineering practice, such as increased use of signals. However, since this practice is likely to be cost-prohibitive at all but the highest crash sites, a suggested benefit of restricting certain, high-risk older drivers to travel on familiar routes should be evaluated, in controlled studies wherever permissible.
The difficulty remains in identifying who the high-risk older drivers are. Test reliability problems plague many procedures for measuring driving-related impairments. Poor results on component measures of driving ability may mask overall competence for the driving task, at least under a given set of conditions, by older persons who apply a variety of compensatory strategies. Finally, practical limitations in the time, expense, and/or complexity of any assessment procedures considered for large-scale implementation among the older population suggest that the greatest contribution to improved safety may result from measures designed to identify only the most clear and profound levels of diminished functional capability.