Classifying Memory Clinic Patients Using the Digital Clock Drawing TestTM, Mini‐Mental State Examination, and Philadelphia (repeatable) Verbal Learning Test

Abstract

AbstractBackgroundDigital cognitive assessments (DCA) enable quick, sensitive, and reliable assessment of neurocognitive operations that underlie dementia and mild cognitive impairment (MCI). We examined how well the DCTclockTM DCA, Mini‐Mental State Examination (MMSE), and Philadelphia (repeatable) Veral Learning Test (PrVLT) can classify memory clinic patients by cognitive impairment profile.Method106 serial referrals to an outpatient memory clinic were assessed with the DCTclock, MMSE, and P(r)VLT. Using statistical criteria for subtle cognitive impairment (SCI, Edmonds et al., 2015) and MCI (Jak et al., 2008), patients were classified into cognitively normal (CN, n = 21), SCI (n = 33), amnestic MCI (aMCI, n = 22), and dysexecutive/mixed MCI (d/mMCI, n = 30) groups. A series of stepwise logistic regression analyses were conducted with CN participants as the criterion group.ResultIn the first analysis (Table 1), P(r)VLT‐recognition (X2= 24.39, p<0.001) entered the model first followed by the DCTclock Score (X2= 16.68, p<0.001). P(r)VLT‐recognition classified aMCI patients (Wald = 11.21, p<0.001). DCTclock (Wald = 8.97, p<0.001) and P(r)VLT‐recognition (Wald = 4.06, p<0.001) classified d/mMCI patients. In the second analysis the four command DCTclock composites were analyzed, excluding the MMSE and P(r)VLT‐recognition scores, Spatial Reasoning classified patients into their respective groups (SCI, Wald = 3.68, p<0.055; aMCI, Wald = 8.84, p<0.003; d/mMCI, Wald = 19.12 p<0.001). Third, analyzing DCTclock command Spatial Reasoning, MMSE, and P(r)VLT‐recognition, the P(r)VLT score, again, entered first (X2= 24.39, p<0.001) followed by DCTclock Spatial Reasoning (X2= 18.95, p<0.001). DCTclock Spatial Reasoning classified SCI patients (Wald = 4.66, p<0.031), P(r)VLT‐recognition classified aMCI patients (Wald = 9.11, p<0.003), and both DCTclock command Spatial Reasoning (Wald = 10.56, p<0.001) and P(r)VLT‐recognition (Wald = 3.32, p<.068) classified d/mMCI patients.ConclusionUnlike the MMSE that takes 10‐15 minutes to administer and is scored subjectively, all DCTclock data are digitally obtained and automatically scored, requiring no examiner interface for administration and interpretation, thus obviating clinical judgment, ensuring inter‐rater reliability, and enabling more efficient cognitive assessments. In this series of logistic regression analyses, a combination of DCTclock and P(r)VLT clearly outperformed the MMSE in classifying non‐demented memory clinic patients into cognitively normal, subtle cognitive impairment, and MCI subtypes. These results support the efficacy of DCAs in clinical trials and to screen for emergent cognitive impairment in primary care and specialty medical practices.