Twelve studies (5,203 participants, range 77 to 2,043) met the inclusion criteria. Of the 12 studies, nine met all 14 of the QUADAS criteria; three studies failed one or two of the criteria (criteria not reported). Three studies had losses to follow-up of 40% or more. Duration of follow-up ranged from six months to three years.
Three tools had high specificities: the Mini Mental State Examination (MMSE) at 96%, the standardised MMSE (S-MMSE) at 100%, and the Cambridge Cognitive Examination (CAMCOG) at 96%. The tools with the lowest specificities were the Montreal Cognitive Assessment (MoCA) at 50%, the Addenbrooke’s Cognitive Examination (ACE) at 43%, and the combination of the MMSE plus Isaacs Set Test (IST) plus Benton’s Visual Retention Test (BVRT) at 52.2%. Specificities were improved with different cut-off values for ACE (to 83%) and the combination of tests (to 91.3%).
Tests with the highest sensitivities included: the Revised Hasegawa’s Dementia Scale (HDS-R) at 92%. the ACE at 100%, the MoCA at 94%, the Chinese Abbreviated Mild Cognitive Impairment Test (CAMCI) at 83.4%, the Cognitive Capacity Screening Examination (CCSE) at 88.1%, and the combination of the MMSE plus IST plus BVRT at 90.8%. Sensitivities of these tests did decrease with some cut-off values or in certain populations (details given). Tests with the lowest sensitivities included the MMSE at 31%, the S-MMSE at 14%, and the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) at 41%.
Results for predictive power, cognitive domain coverage and feasibility of use were also reported.