Screening and Case-Finding Tools for the Detection of Dementia. Part II: Evidence-Based Meta-Analysis of Single-Domain Tests

Abstract

To evaluate the diagnostic validity of single-domain cognitive tests for detecting dementia. These methods were defined as those that focus mainly on one domain of cognitive function such as orientation, memory, or executive function. A meta-analysis of robust studies was conducted. There were 15 categories of single-domain methods assessed in 45 analyses, including 27 performed head-to-head against the Mini-Mental State Examination (MMSE). Thirteen studies took place in community settings, 9 in primary care, and 23 in specialist settings. In community settings, single-domain cognitive tests helped detect 64.2% of all dementias. In this setting, specificity (Sp) was 84.9%, positive predictive value (PPV) was 57.1%, and negative predictive value (NPV) was 88.3%. This was significantly less accurate than the MMSE itself. The optimal individual method was the memory impairment screen (MIS). In primary care, the sensitivity (Se) and specificity of single-domain tests were 69.5% and 82.5%, respectively. The PPV and NPV were 36.5% and 95.8%. Considered together, these methods were less specific but equally sensitive as the MMSE when applied in primary care. Here, the most successful methods were the selective reminding test (SRT) and clock drawing test, the latter having data from independent samples. In specialist settings, single-domain methods had a Se of 76.6%, a Sp of 81.9%, a PPV 80.8%, and a NPV 74.9%. This represented almost equivalent accuracy to the MMSE. The optimal method for specialist settings (based on accuracy alone) was the memory alteration test. Brief single-domain methods offer diagnostic performance for detection of dementia that is surprisingly close to that offered by cognitive batteries such as the MMSE. As a method of screening or as part of a diagnostic algorithm, brief single-domain tests may be an efficient first step in identifying cognitive impairment.