The following is an edited form of an address by Dr. Allen Dobbs to a Toronto symposium sponsored by Alzheimer Toronto and the Rehabilitation Institute of Toronto earlier this year. Dr. Dobbs is a pychologist and Professor Emeritus at the University of Alberta.

"30 to 50% of dementia patients who drive have a crash within a few years of diagnosis. 80% of those continue to drive, and 20% have at least one more crash. At the same time, there are a significant number of dementia patients who retain the competence to drive through the mild and even moderate stages of dementia. So our question then is who's who?; how do we tell which are competent to drive and those who are truly dangerous? If we couple that with the knowledge that driving is one of the major ways older people maintain independence and mobility, it becomes clear we have to answer these questions appropriately.

On one side we have mobility and independence, on the other we have potential disasters. So that was our goal, and its context. How do we evaluate? One goal of our research was to develop a road test that would effectively identify people who were incompetent, and produce the data that show why.

A major premise was that not all errors of driving indicate declining competence, that some errors might be the kind that even competent drivers make. This was a very strong implication, that any justifiable evaluation must identify and penalize drivers only for errors that can be documented as a signal of declining competence. So how would we know? Well, you take one group of cognitively-impaired (mostly Alzheimer patients) drivers, another group of healthy drivers of similar age and a third group of younger drivers. You put them all in a car, let them drive around and study their behavior. Any common behavior you ignore; what you look for is behavior that is specific to Alzheimer patients. That's what we did. Later, we examined whether an on-road test was enough, and I'll show you why it isn't": (At this point, Dr. Dobbs shows a series of video clips of drivers who, at times, did complete U-turns when asked to turn left on city streets, drove against the traffic on a freeway or drove onto the median. These were actual scenes taped during a series of tests using cameras and an evaluator sitting next to the driver. Ironically, much of the Toronto audience watching these potentially-horrific scenes laughed.)

"Obviously, too many of these drivers are unsafe, really unsafe. They are just too dangerous to be on the roads. So we felt we had to develop some kind if a screen beforehand: one that would predict accurately which drivers were likely to fail and which were likely to pass. Thus, we would reduce the numbers we needed to take onto the road while at the same time increase the cost-effectiveness (ie: fewer drivers, less cost).

The bad drivers we chose for the first stage, 176 in total, were all referred by a physician because they were cognitively-impaired and still driving. We also had a group of 70 non-impaired volunteers who were roughly the same median age as the first group, and a third group of men and woman aged between 30 and 40 years. All were subjected to clinical evaluation: remedial medical and neuropsychology. The results were computerized, and our goal was to develop something short, an evaluation that would not take inordinate amounts of time. We put the participants through a road test involving 37 separate manoeuvers: right turns, left turns, things like that. It was a closed course with conditions similar to those known to involve accidents among older people. We ended up with a slew of driving errors, which we categorized into 13 sectors: signaling, positioning, etc. We also had one category which we labeled Near Catastrophic, things that would result in a crash had the evaluator not taken control, or had traffic not adjusted. We then compared the three groups for statistical differences. That led to three categories of errors that were really defining in terms of what we were looking for.

Those three categories were: Errors That Do Not Count: these were common to all groups: things like errors at stop signs, improper signaling etc. Discriminating Errors: these were the errors that saw the younger people make a few, older people made more and the demented people make the most. We could identify a cut point beyond which no normal person would be within the range. So we could say "if you're beyond this range, you should not be on the road. Criterion Errors: These were really hazardous errors.

I can assure you no normal person, young or old, would make these errors. Things like driving the wrong way on the freeway. I can assure you no normal person would commit these errors during a driving evaluation. As a result, we were able to devise a driving evaluation based, for the first time, on data. How do we proceed? We went back and looked at our road course and plotted where every error occurred. So we could eliminate those areas that produced little problem, and highlight other areas that brought out very interesting errors (including those that no normal person would make, thus signaling cognitive impairment). We wanted to mimic those kinds of conditions. At the end of that process, we had a pretty good competence screen.

We could develop a battery of tests that predict driving competence. We then devised a 30-minute, computerized test that was 94% accurate (based on 376 new cognitively-impaired people), while at the same time reducing the number of people who need a road test. The advantages of this test is that it is empirically based. It allows us to say with accuracy that healthy older people can pass the test, and that's something seniors groups really like: people are not penalized for errors that all drivers make. The road design is one based on data, and the criterion for an unsafe driver is empirically-based, not one that's a best guess. Right now, the Alberta Traffic Safety Board (Licensing Bureau) have accepted the evaluation as their standard for road testing."