Clinical Utility of Office-Based Cognitive Predictors of Fitness to Drive in Persons with Dementia: A Systematic Review - Molnar, Patel, Marshall, Man-Son-Hing & Wilson - 2006 - Article

The population of developed countries continues to age, so the prevalence of age-associated disease will increase. This also means that the number of persons with dementia who are actively driving is expected to increase. Since the late 1960s, the matter of driving in people with dementia has become a topic of growing societal concern. While there are some studies that state that there are no increased risk of motor vehicle crashes (MVC) in people with dementia, most studies show that, as a group, people with dementia have a 2.5 to 4.7 times greater risk of MVC than control groups. There might not be a single precise and generalized risk estimate achievable, because the estimates will vary because of different sample sizes, different clinical populations, varying levels of driving exposure, different definitions of MVC, different means of measuring MVC rate and whether the MVC rate is calculated per distance driven or per driver. However, the fundamental finding of many studies is that people with dementia are at a higher risk of experiencing a MVC than the population average. Therefore, early driving researchers recommended that all patients diagnosed with dementia cease driving. Arguments that justify this position are the variability within the category of mild dementia, the unpredictability of progression rates, the different effects on various types of dementia on fitness to drive and the increased propensity of people with dementia to experience sudden acute confusional states. Basically, the arguments regarding dementia and driving focus on the unpredictability of progression to unsafe states.

Some people have argued that the conservative approach of removing the licences of all people diagnosed with dementia, presents several problems while not resolving the safety concern. This approach penalizes patients who have been given a diagnosis of dementia and this may discourage people with possible dementia from seeking a diagnosis. Also, some people have reversible disorders and there are sometimes uncertainties with diagnoses, so the licence might be removed from people who have reversible disorders rather than progressive dementia. Some studies have shown that many people with dementia may still be safe to drive. One study showed that the majority of people with dementia who were actively driving had not been involved in MVCs. However, certain biases (recall bias) might limit this study. It is often hard to assess driving exposure and that limits database research. There are some medical associations that recognize that some people with mild dementia may be safe to continue driving. They do not recommend reporting based on diagnosis, but on individual functional assessment of persons with mild dementia. There are guidelines that provide recommendations for a functional assessment of fitness to drive, like the American Medical Association’s (AMA) document called ‘Physician’s Guide to Assessing and Counseling Older Drivers.’ This guide assists clinicians by providing tools to screen all drivers rather than focusing on the evaluation of dementia exclusively. All screening tests must be as rigorously evaluated and validated as more-complex assessment procedures.

For more evidence-based guidelines for clinicians, a systematic review of the literature, from a clinical perspective, is required. In this article, the evidence for future versions of guidelines will be examined via a systematic literature review of studies that have attempted to answer the questions: ‘Which office-based cognitive tests have been analysed in a manner that permits their immediate application in frontline clinical settings in the assessment of fitness to drive in persons with dementia?’ this review also looks whether tests are associated with MVCs, whether evidence-based and validated cu-toff scores are provided and whether the standard psychometric properties of the cut-off scores are reported. Because there is much variability in study designs, definitions of MVC and outcome measures, a meta-analysis is not possible.

Method

The focus of the review is on in-office assessment and tests that require specialized equipment not readily available in usual healthcare settings were not included. When it was not certain if equipment was required, the test was included in the review. Supplementary skills that are related to driving, like strategic and operational abilities were not examined. Complementary areas, like sensory functions, sociodemographic factors and behavioural factors were not examined. Because of ethical considerations, randomized trials can’t be employed. To collect cognitive tests, all cohort, correlation and case-control studies that tried to answer the main question posed were included. Only studies that used accepted diagnostic criteria for dementia (from the DSM), or Alzheimer’s disease criteria from the National Institute of Neurological and Communicative Disorders and Stroke- Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) were selected. Also, only articles published in English were considered. The literature was searched in a couple of databases. First, the researchers looked at the title and selected articles were looked at more closely to see whether they would fit in the review. The texts of 164 articles were reviewed, with 16 articles meeting the inclusion criteria. The primary outcome was history of crash for six studies, stimulator test results for four studies and on-road assessment for six studies. A great amount of inconsistency regarding the association between test results and driving outcome measures was found. The outcome measure used appeared to have an influence on findings. Positive associations of certain cognitive deficits were with the ability to drive as determined by performance-based testing protocols rather than with real-life performance.

Results

The association between each test and MVC and the findings within each domain show a great amount of variability. These are probably the result of a number of methodological factors, like different sample sizes, multiple predictor variables, different outcome measures employed and performance measures. The outcome measures have limitations and these also contribute to the variable results. In table 4 on page 1819 of the original article, you can read all the limitations of every construct studied. This study shows a significant research gap that represents a barrier to the creation of evidence-based fitness-to-drive guidelines. The results with respect to the association between individual tests and MVCs were variable, but the consistent findings of these studies was that none of the included studies were analysed in a manner that could generate clinically usable tools for front-line clinicians. The studies also don’t give data required for the development of evidence-based fitness to drive guidelines. Clinicians need to label a patients as a safe or unsafe driver. This task requires that tests have evidence-based cut-off scores that suggest a course of action. However, all the studies (except for one) reviewed didn’t look at cut-off scores, but at association. Knowing that a certain cognitive test is associated with risk of MVC is helpful in selecting tests that merit further study. However, when there are no cut-off scores, the association is of little practical utility. The study that did provide cut-off scores did not provide psychometric properties, like sensitivity or specificity. It is difficult to find cut-off scores. The cognitive scores of people who are fit to drive and those who are unfit to drive probably have overlapping distributions. Because of the overlapping test scores of safe and unsafe drivers, it is unlikely that a single clinically usable cut-off score will be found for most tests. This may be the reason why some studies do not report cut-off scores.

Multivariate analytic techniques are the conventional analytic approaches employed in the field of research. These techniques are based on the assumption that all needed information is available at one point in time. This assumption may be valid in subspecialty settings, but the techniques do not model routine front-line clinical decision-making. Multivariate equations are never employed in front-line clinical practice. In clinical practice, decisions are first made based on minimal information. Other tests are applied to patients whose status could not be determined via the initial information. This is a more efficient approach to maximize ability to diagnose while minimizing cost and morbidity associated with more-expensive and invasive tests. If driving researchers do develop tools to assist front-line clinicians, then they should also look beyond the conventional multivariate techniques to include cut-off based approaches that better model front-line clinical decision-making. So, research needs to move on finding out more about cut-off scores and deciding what values to aim for with cut-off scores. There is also not much research on the determination of risk threshold. How much risk is too much risk? Without a risk threshold, clinicians will have to base their decisions of safe or unsafe drivers on their personal opinion. A risk threshold is needed to standardize the practice. Standardized risk thresholds will set the criteria that cut-off scores are to meet and researchers can compare different studies more easily in that way. Another thing that future research can examine, is the pattern of cognitive deficits and specific types of deficits on cognitive tests and how they relate to the fitness to drive. Researchers should also be encourage to perform systematic reviews of cognitive impairment and driving. Future research should also have adequate sample size to avoid error.