The timely recognition of dementia is the prerequisite for adequate information, treatment, and care. Nevertheless, dementia is known to be considerably underdiagnosed; even in high-income countries with advanced medical care systems about 50% to 80% of people with dementia (PWD) are not formally diagnosed [1,2]. There are approaches to improve the recognition of dementia, such as the proactive “dementia case finding scheme” that was initiated by the government of the United Kingdom  or the “Annual Wellness Visit” for Medicare enrollees in the United States which includes the detection of any cognitive impairment . However, the best practice for the identification of dementia in primary care has not yet been established. Previous studies showed that the use of structured screening instruments improves the identification of cognitive impairment in primary care and that the screening for dementia increases diagnosis rates [5–7]. Nevertheless, routine screening is controversially discussed and not recommended in respective dementia guidelines because there is still a lack of evidence that patients benefit from it 69 8 [8–12]. Arguments against routine screening include the risk of receiving a false-positive diagnosis of dementia after a positive screening outcome; the cause of anxiety or depression among positively screened subjects; unnecessary examinations and treatments; the diversion of resources that would better be used to care for real dementia cases; or the danger that older patients will avoid visiting their general practitioner (GP) because they fear to be diagnosed with dementia [13–16]. Therefore, routine screening of asymptomatic patients is not seen as the favorable solution to improve the recognition of patients with dementia in primary care. It has been suggested that the case-finding of dementia should focus on patients presenting with cognitive complaints [14,15].
Subjective memory impairment (SMI) may represent the first symptomatic manifestation of Alzheimer\’s disease and SMI and related worries have been identified as risk factors for the incidence of dementia in people without objective cognitive impairment [17–19]. However, the diagnostic value of self-reported cognitive impairment for prevalent dementia seems to be limited for several reasons: SMI is associated with depression [20–22]; cognitively healthy older persons frequently complain about memory impairment [23,24]; and PWD are often not aware of their memory problems [25,26].
Mitchell  conducted a meta-analysis of the diagnostic value of subjective memory complaints for manifest dementia in community samples with a low prevalence of dementia and found a positive predictive value (PPV) of 19% and a negative predictive value (NPV) of 94%. Mitchell concludes that the absence of subjective memory complaints may be a reasonable method of excluding dementia and could be incorporated into short screening programs in settings with low prevalence of dementia. However, to our knowledge this assumption has not yet been validated in clinical settings. This study aims to determine whether self-reported SMI or SMI-related worries could be used as a valid criteria to decide if an elderly primary care patient should be screened for dementia.
Our analyses showed that the sensitivity of SMI was just 54%. This means that 46% of the primary care patients screened positive for dementia did not report SMI before the screening. Thus almost half of the patients with cognitive impairment would have been overlooked if the presence of SMI would have been the precondition for performing the structured cognitive test. In addition, both the PPV (19%) and the UI+ (0.10; interpreted as “poor” rule-in-accuracy) of SMI were very low. Mitchell  reported similar results with a meta-analytic pooled sensitivity of 43%, a PPV of 19%, and an UI+ of 0.08 and concluded that “subjective memory complaints should not be relied on for case-finding”. These findings are in line with the results of previous studies that showed that people with dementia are often not aware of their memory problems [25,26].