How to choose the most appropriate cognitive test to evaluate cognitive complaints in primary care

When a patient complains but the GP considers cognitive impairment to be not likely, the prior probability that this patient has cognitive impairment is low and the chance this patient has dementia will be even lower [4]. The main objective of a cognitive test in this situation is to rule out cognitive impairment, in particular MCI. A test should have a high negative predictive value (NPV) and should preferably be brief. A high positive predictive value (PPV) is less relevant if one aims to rule out a condition, as a low PPV can be amended by performing an additional test in case of a positive test result. For this situation, we only considered tests that have been studied for MCI.

This is the most challenging diagnostic situation, the “grey zone”. When the GP considers cognitive impairment to be possible, but ADL appears to be preserved, the prior probability that the patient has MCI, or possibly even dementia is substantial [4]. The main objective of a cognitive test in this situation is to distinguish between presence or absence of cognitive impairment. A cognitive test in this situation should therefore be able to detect MCI and dementia in a population with a moderately high prevalence of cognitive impairment. We may assume that a test validated for MCI with an adequate NPV, will also detect dementia. Therefore, we considered tests that have been studied for MCI only, or MCI and dementia. We prioritised a high NPV above a high PPV to avoid false reassurance.

When the GP considers the likelihood of cognitive impairment to be high and ADL appears to be affected, the prior probability that this patient has dementia is high [4]. A cognitive test in this situation should therefore be able to detect dementia in a population with a high prevalence of cognitive impairment. For this situation, we only considered tests that have been studied for dementia. The main objective of a test in this situation is to confirm that the patient has dementia; a test with a high PPV for dementia is therefore preferred.

As shown in Table 1, both the clock-drawing test [18] and the Montreal cognitive assessment (MoCA) [19] have a high (≥ 89%) NPV and a moderate PPV (≤50%) in populations with relatively low prevalence rates of MCI (14–24%). Taking into account their comparable diagnostic accuracy, the short administration time of the clock-drawing test (1–3 min) relatively to the MoCA (10 min), we selected the clock-drawing test for our algorithm. The clock-drawing test assesses multiple aspects of cognitive functioning, in particular visuospatial and praxis abilities. In contrast, the MoCA contains multiple subtests that tap into different cognitive domains and can thus provide some more information on the actual nature of the cognitive impairment.

As shown in Tables 1 and 2, all tests that have been studied for MCI only, or cognitive impairment, have limited PPVs (≤71%), with the exception of study populations in which MCI or cognitive impairment is highly prevalent (≥50%). The MoCA has the most favorable NPV (≥ 94%) for both MCI and cognitive impairment overall and was therefore selected for our algorithm.

Table 3 demonstrates that the Mental Status Questionnaire, the Short Portable Mental Status Questionnaire and the Memory Impairment Screen were only investigated in study populations with a prevalence of dementia ≤18% and it is therefore unclear if they are suitable in a situation with a high prior probability of dementia. The Abbreviated Mental Test and the Mini-Cog were both studied twice, once in a population with a very low prevalence of dementia (3% and 4% respectively) and once in a population with a high prevalence of dementia (29% and 40% respectively). In populations with a high prevalence of dementia the PPV of both tests was 71%. The MMSE (cut-off <24) has a comparable PPV (73%) in a population with a dementia prevalence of 28%. The NPV of all tests - with the exception of the Abbreviated Mental Test - was above 90%. In conclusion, both the Mini-cog and the MMSE with a cut-off <24 have favourable test features for this situation. Since the MMSE [20] is most frequently studied and well known, we selected this test as most suitable for our algorithm.

The starting point of the algorithm (1.1) is cognitive complaints expressed by the patient or a relative, or suspicion of cognitive impairment by the GP. In the evaluation of cognitive complaints, the mode of onset (1.2) provides essential guidance. In MCI and dementia, which is mostly caused by neurodegenerative or vascular pathologies, cognitive impairment is acquired and has a slowly progressive onset. This algorithm is not applicable to cognitive symptoms that develop within days or weeks. In that situation, other diagnoses such as a delirium or other neurological conditions are more likely.

History taking and an informant interview are fundamental in a cognitive evaluation [2]. Concerns expressed by a close informant are generally even more predictive of cognitive impairment than self-reported symptoms [2]. An informant interview is preferably performed with a close informant separately from the patient. If an informant is not available and diagnostic uncertainty persists after the initial visit, the patient should bring an informant to a follow-up visit. The following topics should be addressed:

Based on the previous steps the GP can estimate the likelihood that the patient has cognitive impairment and choose the most suitable cognitive test (Fig. 1). If according to the GP the likelihood that the patients has cognitive impairment is very low or very high, it may well be that none of the cognitive tests are of added value. Not using any cognitive test could then be a good option.

The steps in the proposed algorithm will guide the GP towards the most probable diagnosis (Fig. 1). However, if there is a mismatch between the findings of history taking and the test, the results need to be reconsidered. It is important to perform an informant interview if not done previously and to consider alternative diagnoses. If uncertainty persists, the GP may decide to re-evaluate the patient in 6–12 months or to refer to a specialist for a more comprehensive cognitive assessment.