Should doctors screen their elderly patients for cognitive impairment?
Because of the high prevalence and social costs of dementias in late life and the emergence of useful therapies, a growing consensus favours cognitive screening as part of routine primary care of the elderly (1–4). The goals of cognitive screening in primary care differ considerably from those of dementia epidemiology, and methods for evaluating its utility will also differ. Routine dementia screening in primary care could achieve several useful objectives in addition to dementia detection: it could sensitize primary care physicians to the possibility of declining cognition in their older patients, accelerate translation of research advances into actual practice, promote development of quality standards for dementia care across practice sites and styles, and encourage design of proactive strategies for population-based health care of dementia patients and their families (5,6).
Screening will become increasingly important as the elderly population grows, because dementia prevalence dramatically increases with age (7), and most dementia detection will occur at the primary rather than specialty care level. To meet the unique objectives of screening in primary care, instruments that are brief, reliable, and simple to use are of primary importance. This paper examines various brief dementia screens with the goal of identifying those that 1) best classify subjects with, and without, cognitive impairment, 2) are most suitable for primary care, and 3) are likely to be used routinely by physicians.
Is screening being done? If not, why not?
Dementia screening in primary care settings is infrequent, and questions asked about cognition during a primary care visit generally follow no standard format (8). Several studies conducted in primary care settings over the past 20 years document high rates of missed dementia diagnoses (5). More recently, Valcour and others found that dementia was missed in 67% of all affected patients and in 91% of cases when impairment was mild (9). General practitioner surveys have discovered that only 24% of Canadian (10) and 39% of Australian (3) doctors surveyed routinely screen for dementia using any recognized method. Many of these physicians are familiar with standard dementia screens—the best known being the Mini-Mental State Examination (MMSE) (11)—but consider them too long and too difficult to interpret. Some believe that the process of dementia screening itself would be threatening to patients and are therefore reluctant to do it. However, if a brief and simple instrument were proven effective, 93% of physicians surveyed said they would use it (10).
Administration time is a critical determinant of the acceptability of routine dementia screening. What is too long? Increasingly scarce contact time between doctors and patients dictates that screening should be possible in a fraction of the time allowed for a typical office visit, or perhaps even be done by office staff while patients wait for appointments (5). The average length of a primary care visit varies from 7.5 minutes in the UK (12) to 19 minutes in the US (13). Screens that can be administered in 5 minutes or less would seem the most likely to be useable in primary care. Given that the MMSE typically requires 5 to 10 minutes to give (although, as for all screens, it requires less time for robust, cognitively intact younger elderly and more time for very old individuals with dementia) and is still underutilized by doctors, we therefore emphasize instruments completed in 10 minutes or less.
Contrasting Approaches to Dementia Screening
Dementia screening tools fall into 3 main categories: cognitive tests administered to patients, questions about cognition and function asked of proxy informants, and functional assessments that use direct observation-of-test tasks. Cognitive tests provide a snapshot of current cognitive performance, while informant-rated screens can provide a longitudinal view of changes in cognition-dependent behaviours over time. Some instruments combine elements of both approaches in a single composite tool. Direct assessments of function observe the patient’s ability to perform specific tasks, usually related to activities of daily living. The Medication Management Test (14), which measures a patient’s ability to learn and use a hypothetical daily medication regimen, is an example of this approach that has been validated against direct cognitive screening. However, functional screens are considerably less developed than cognitive- and informant-based approaches and will not be further addressed here.
Conceptual and Methodological Issues in Evaluating Screening Instruments
Several factors influence the effectiveness of screening instruments. Therefore , results should always be interpreted with the study design and goals in mind. In general, screening instruments perform best when the target sample has a high proportion of individuals with dementia and the severity of impairment in cases is high. They perform worst in samples containing infrequent cases of mild dementia. This means that studies examining different tests in samples dissimilar along these dimensions will not be directly comparable. Typical ways of reporting performance, such as sensitivity, specificity, and predictive values of positive and negative tests, are not fixed properties of test instruments but features of how they perform under specific applications to specific study samples.
Tests should be validated against a diagnostic standard applied equally to subjects with, and without, dementia and to those with mild, ambiguous, and questionable impairments. If the entire sample is not subjected to stringent neuropsychiatric evaluation (for example, if normal subjects are classified only by a screening test and not approached with the same diagnostic rigour as are subjects with dementia), individuals with mild cognitive impairment could be inadvertently classified as normal. If screening tests are examined in clinical or convenience samples that maximize differences between normal subjects and those with dementia (a common and appropriate initial step in instrument development), results will not constitute a definitive performance test in “real-world” application. If studies are conducted on samples that restrict subjects’ education and language (such as samples including only those with at least an 8th grade education who speak English), these sampling constraints could hinder generalizability.
Ideally, tests should be validated in population-based samples that represent the target groups in which a screen will ultimately be used. A reasonable compromise with this gold standard calls for efforts to validate instruments in samples representing as broad a range as possible of education levels and ethnolinguistic and cultural backgrounds. Examining heterogeneous samples is particularly important in regard to screens considered for use in primary care settings. This is because minority elders are the fastest-growing population of older adults in developed countries such as the US (15) and pose the greatest challenges for dementia screening.
Tests can also be validated by comparing them with a known reference standard, such as the MMSE, on the same population. Stuss states that cognitive screening tests must be formally compared to effectively choose between them (16). (Various studies that have done this [16–18] favour comparisons using receiver-operating characteristics [ROC], because these are better measures of performance when cut-off points are not standard or do not apply to a given sample.)
Archives in English | Archives
RCP en français