Diagnostic accuracy of virtual cognitive assessment and testing: Systematic review and meta-analysis

Background/Objectives Virtual (i.e., telephone or videoconference) care was broadly implemented because of the COVID-19 pandemic. Our objectives were to compare the diagnostic accuracy of virtual to in-person cognitive assessments and tests and barriers to virtual cognitive assessment implementation. Design Systematic review and meta-analysis. Setting MEDLINE, EMBASE, CDSR, CENTRAL, PsycINFO, and gray literature (inception to April 1, 2020). Participants and interventions Studies describing the accuracy or reliability of virtual compared with in-person cognitive assessments (i.e., reference standard) for diagnosing dementia or mild cognitive impairment (MCI), identifying virtual cognitive test cutoffs suggestive of dementia or MCI, or describing correlations between virtual and in-person cognitive test scores in adults. Measurements Reviewer pairs independently conducted study screening, data abstraction, and risk of bias appraisal. Results Our systematic review included 121 studies (15,832 patients). Two studies demonstrated that virtual cognitive assessments could diagnose dementia with good reliability compared with in-person cognitive assessments: weighted kappa 0.51 (95% confidence interval [CI] 0.41-0.62) and 0.63 (95% CI 0.4-0.9), respectively. Videoconference-based cognitive assessments were 100% sensitive and specific for diagnosing dementia compared with in-person cognitive assessments in a third study. No studies compared telephone with in-person cognitive assessment accuracy. The Telephone Interview for Cognitive Status (TICS; maximum score 41) and modified TICS (maximum score 50) were the only virtual cognitive tests compared with in-person cognitive assessments in >2 studies with extractable data for meta-analysis. The optimal TICS cutoff suggestive of dementia ranged from 22 to 33, but it was 28 or 30 when testing was conducted in English (10 studies; 1673 patients). Optimal modified TICS cutoffs suggestive of MCI ranged from 28 to 31 (3 studies; 525 patients). Sensory impairment was the most often voiced condition affecting assessment. Conclusion Although there is substantial evidence supporting virtual cognitive assessment and testing, we identified critical gaps in diagnostic certainty.

Automated summary

Virtual cognitive assessments showed good reliability in diagnosing dementia, with videoconference-based assessments having 100% sensitivity and specificity for dementia diagnosis. However, there are critical gaps in diagnostic certainty, indicating the need for further research in this area.