BRAINSTORMING SESSION (continued)

Topic Area 3:  What are the Most Practical and Valid Ways of Assessing the Impact of Medication Usage on Actual Driving Performance?

Discussion Summary

• Guided by the Literature Review, this discussion specifically emphasized driving performance as an outcome, as crashes are rare events.  The focus was on identifying the best ways to test driving competence, using measurable behaviors, as opposed to epidemiological studies where crashes are the outcome.

• A panelist questioned the practicality of using OT/CDRSs to conduct on-road evaluations, due to the limited number and availability of these specialists in the U.S.  

• Another panelist suggested making use of trained driving instructors for conducting in-car evaluations, to overcome the problem of limited numbers of OT/CDRSs. 

• OT/CDRS panelists commented that, in any study designed to measure the effect of a medication on driving performance (either an improvement or a decrement), it would be desirable to conduct a driving evaluation prior to the administration of medications; this would require contact with prescribing physicians, which reduces the practicality of this method. 

• A physician noted that judging the best method of determining drug effects on driving performance depends on the question being asked.  To investigate whether an underlying condition believed to be responsible for driving difficulties can be successfully treated with medications (i.e., to improve driving performance) may require a different approach than testing an hypothesized negative effect of drugs on driving.  

• Instrumenting peoples’ vehicles could be a way of collecting more data, more quickly and efficiently.  However, because there is no driving evaluator to intervene if the driver makes a critical or hazardous error, this is ethical only if people are being asked to drive as they usually do under their current medication regime.  It would not be acceptable as an experimental approach where new medicines are being introduced.

• An OT offered that it could also be useful to instrument an evaluator’s car to pick up subtle effects related to medication; for example, a medication may have a particular physical effect, or otherwise impact the person’s capacity to drive in a manner that is difficult to observe directly.

• Participants suggested that it might be practical to use GPS to determine how people who are driving their own vehicles find their way from point to point, while acknowledging that many unsafe behaviors (like stopping in the middle of an intersection to read signs) would not be picked up with a GPS.  An outboard camera or an evaluator would be required to determine this. 

• Adding cameras to instrument a person’s own car could also be used to record what was happening outside of the vehicle, to establish a context for the driver’s behavior.  There remains a caution with any research protocol that requires someone to drive in an unfamiliar area without a driving evaluator, however, as this introduces risk that would not otherwise be part of the driver’s daily routine.

• If a person has adapted to medication, it may take a long time for any adverse driving events to show up, if at all.  Exposure to a wide array of unexpected events and situations would be required to determine all safety consequences of medication use; this suggests testing on a closed course utilizing a driving evaluator or OT with a properly equipped car (to intervene if necessary), or testing in a simulator. 

• The executive component of driving is the most important thing to test when researching the effects of polypharmacy, according to the driving evaluators.  A clinical assessment of cognitive function should be done before going on-the-road; without this information, it will be more difficult to attribute any effects on driving performance to medication use. 

• Ideally, a pre-test of driving functioning should be conducted prior to people taking their medications.  Or, a study could require people to stop taking their medications, to obtain a baseline measure of performance. 

• Because it is highly unlikely that peoples’ medication regimes could be changed, research designs for future work in this area could benefit from multivariate analysis techniques to help explain the contributions of medications, medical conditions, age, cognitive function score, etc. on driving performance.

• It may be more fruitful to do “data mining” in administrative databases than to conduct an instrumented vehicle study to examine “global issues.”  However, to answer small, specific questions, and/or to test hypotheses generated through prior database analyses, an instrumented vehicle study could be appropriate. A continuing research program on polypharmacy and driving could be designed to first perform “database studies” to determine key medications or conditions.  Then, the combinations of interest can be evaluated experimentally, using the methods favored by participants in this brainstorming session in studies with smaller test samples. 

• A database expert underscored the benefit of using large population databases to identify potential associations or potential risks. This is a relatively low-cost approach.  Other database experts indicated that it is possible to link databases, yielding a great richness of information, although there may be special authorizations required in the Federal sector to do this. 

• A panelist suggested analyzing driving performance data that may already reside in State DMVs, in relation to prescription drug information from a database like Medicare or Medicaid.  It was also noted that, currently, only Illinois and New Hampshire require renewing drivers (over age 75) to take an on-road test.   

• An OT/CDRS perspective is that these professionals (OTs) routinely perform rigorous on-road evaluations.  While such road tests may vary slightly, and it is important to have the inter-rater reliability between OTs, these road tests should be strong candidates for use in NHTSA studies on drugs and driving.

Rating Scale Responses

           Two sets of rating scale responses were solicited to capture the discussion group’s opinions in this topic area, one relating to practicality/reliability/cost-effectiveness of alternative methods and the other relating to older persons’ willingness to participate in research.

In the first case, seven methods of measuring driving performance were considered by the experts participating in the brainstorming session:

1. Closed Course (Controlled Exposure).
2. On-Road, In Traffic (with a Trained Observer).
3. Simulation Level III: Interactive, Computer Graphic Visuals, Full Motion.
4. Simulation Level II: Interactive, Computer Graphic Visuals, Restricted Motion or No Motion.
5. Simulation Level I: Non-Interactive, Computer Graphic and/or Digital Video Visuals, No Motion.
6. Instrumented Vehicle (with Driver’s Own Car).
7. Functional Measures Validated as Crash Predictors.

Summary statistics are presented in Appendix F, for all experts together, and broken out by area of expertise. Results are briefly described below.

With regard to the practicality of one method vs. another and across all experts, the top three measures were: functional measures validated as crash predictors (mean rating = 75), closed course (mean rating = 65), and on-road, in traffic (mean rating = 63).  Level III simulation received the lowest rating (mean rating = 39).  Ranked similarly and falling midway between the best and worst methods were the remaining two simulation levels and the instrumented vehicle method (ratings ranged between 54 and 58).  Physicians and pharmacologists rated functional measures as the most practical method (mean rating = 75), followed by simulation level I (mean rating = 69), and on-road/in traffic measures (mean rating = 63).  Driving evaluators tied on-road testing and functional measures as the highest (mean rating = 72), followed by closed course evaluations (mean rating = 70).  They rated the three simulation measures as the lowest (ranging from 28 to 38), with the higher-fidelity methods rated poorer than the lower-fidelity methods. The behavioral researchers rated the functional measures best (mean rating = 78), followed by simulation level II (mean rating = 73), and the closed-course method (mean rating = 68).  Interestingly, for the instrumented vehicle method, the behavioral researchers provided the lowest rating of all the expert types (mean rating = 25) as well as the simulation level III method (mean rating = 20). The database experts rated the functional measures highest (mean rating = 75), followed by the instrumented vehicle method (mean rating = 68), and the closed course method (mean rating = 65).  

With regard to reliability, the highest-rated method across expert types was the instrumented vehicle (mean rating = 72), followed closely by the on-road method (mean rating = 71) and the functional measures (mean rating = 70).  Simulation level I was given the lowest average rating (51).  The same pattern was seen in the ratings of the physician/pharmacist group, with the functional measures and the instrumented vehicle method tied as the highest (mean rating = 76), followed by the on-road method (mean rating = 74).  The driving evaluators grouped the on-road test and the functional measures as the best method (mean rating = 90), followed by the combination of the closed course method and functional measures (mean rating = 75).  Without grouping methods, driving evaluators rated the functional measures as the best method (mean rating = 70), followed by the instrumented vehicle and on-road tests (tied at 63), and simulation level III (mean rating = 62).  The behavioral researchers rated the closed-course method highest (mean rating = 80), followed by the instrumented vehicle method (mean rating = 67), and functional measures (mean rating = 65).  They rated simulation level I as the poorest (mean rating = 28).  The database experts rated the on-road method as the most reliable (mean rating = 85), followed by the instrumented vehicle method (mean rating = 82), and the closed course method (mean rating = 80). 

In terms of cost-effectiveness, the best-rated method across all expert groups was the functional measures method (mean rating = 77), followed by the instrumented vehicle (mean rating = 63), and the closed-course method (mean rating = 61).  The simulation methods were the poorest-rated measures (ratings ranged from 30 to 51), with the high-fidelity methods receiving lower ratings. The physician/pharmacist group followed this pattern.  Driving evaluators rated the functional measures as most cost effective (mean rating = 73), followed by the instrumented vehicle (mean rating = 65), and then the on-road method (mean rating = 63).  The behavioral researchers rated the functional measures and the closed course test as the best methods (mean ratings were 79 and 60, respectively); the next highest measure, the instrumented vehicle, received a much lower rating from this group (44).  The database experts rated the functional measures as the most cost-effective method (mean rating = 73), followed by simulation level I (mean rating = 72) and the instrumented vehicle method (mean rating = 68).

For overall ratings, across all experts, the instrumented vehicle and on-road methods were rated highest (68), followed by the functional measures (64) and the closed course method (59).  Simulation level I was rated lowest (mean rating = 40), followed by simulation level III (mean rating = 43) and simulation level II (mean rating = 45).  The physician/pharmacist group rated the functional measures highest overall (mean rating = 72), followed by the instrumented vehicle (mean rating = 70), and the on-road method (mean rating = 68).  One member of this group gave the highest rating of 95 to the combination of three methods: on-road, instrumented vehicle, and functional measures.  The driving evaluators rated the on-road test as the best overall measure at 85, followed by the instrumented vehicle at 72, and the closed course at 48.  One member of this group gave the pairing of closed course, on-the-road, and functional measures the highest possible rating of 100. The behavioral researchers rated the closed course as the best overall method at 65, followed by functional measures at 62, and the instrumented vehicle method at 56.  The database experts rated the instrumented vehicle as the best overall method at 74, followed by the on-road method at 70, the closed course method at 68, and the functional measures at 67.  The simulation measures ranged from 56 to 62, with the lower-fidelity methods receiving higher scores.

In the second set of ratings during this part of the brainstorming session, panelists were asked to rate the likelihood that older people would be willing to participate in research on medication and driving, as a function of the method used to measure driving performance.  The same seven measures (A-G) above were rated.    Summary statistics are presented in Appendix G, for all experts together and broken out by area of expertise.  Results are briefly described below.

Across all expert groups, the functional measures placed the highest in these ratings, at 69, followed by the instrumented vehicle method at 61, the closed course method at 59, and the on-road method at 58.  The simulation methods received mean ratings ranging from 48 to 54, with the lower-fidelity methods receiving higher ratings.  The physician/pharmacist group rated the functional measures the highest at 63, followed by simulation level I at 62, and the on-road tests at 60.  The remaining measures were not rated as radically different from one another, with simulation level II receiving a rating of 58, instrumented vehicle and on-road evaluations effectively tied at 57, and simulation level III at 54.  The driving evaluators rated the functional measures as the highest (mean rating = 78), followed by closed course and on-road measures tied at 65, instrumented vehicle at 56, and the simulation measures ranging from 40 to 42.  The behavioral researchers rated the functional measures as the highest, at 78; followed by the instrumented vehicle at 60, the on-road method as 57, and the closed course at 56.  The simulator methods ranged from 40 to 50, with the lower-fidelity methods again receiving higher ratings.  The database experts rated the instrumented vehicle method highest at 70;  followed by the functional measures at 65, simulation level I at 57, simulation level II at 54, the closed course and on-road evaluations tied at 53, and simulation level III at 52.