Analyze the reliability, validity, learning effects, and psychometrics of a novel digital Clock Drawing Task (dCDT), and evaluate the accuracy and reliability of its automated feature-analysis.

dCDT is a validated digital Clock Drawing Task (CDT), which can be administered in ~3 minutes, and extracts cognitive- and motor-related features, enabling detection of cognitive impairment (CI). The original dCDT relied on a digitizing pen and specific paper, which can be costly and inefficient. We developed a tablet-based dCDT, DCTclock™, which is administered with a stylus on an iPad, decreasing the cost and improving its accessibility. We previously established the CI-classification equivalence of pen- and tablet-based versions of DCTclock™. Here we evaluate its reliability, psychometrics, and accuracy of clock element classification.

Dataset 1 (N=4963), Dataset 2 (N=170), and Dataset 3 (N=501) including cognitively normal (CN) and CI participants were analyzed for psychometrics. Datasets 1 and 4 (N=432) were analyzed for reliability and learning effects. All datasets were analyzed for automated clock element classification accuracy.

DCTclock scores showed good test-retest reliability in a sample of CN and CI participants (r=0.8) and an independent CN sample (r=0.7). 1% (n=10) and 56% (n=581) of CN participants scored at floor and ceiling, respectively, compared to 40% (n=306) and 17% (n=130) for CI participants. Average 1-week and 3-week score changes were +9.8 (p<0.05) and +0.5 (n.s.), respectively. Overall, 1.56% of clocks were unanalyzable due to missing components. Clock element-classification accuracy was 99.6% and 99.3% for CN and CI groups, respectively.

DCTclock™ is a fast cognitive assessment with good test-retest reliability and high sensitivity and accuracy for detection of cognitive impairment. Automated drawing element classification is highly accurate. DCTclock™ is a robust cognitive-screening tool that enables frequent assessments while maintaining reliability in clinical settings.