The Clock Drawing Test – a simple pencil and paper test – has been used for more than 50 years as a screening tool to differentiate normal individuals from those with cognitive impairment, and has proven useful in helping to diagnose cognitive dysfunction associated with neurological disorders such as Alzheimer’s disease, Parkinson’s disease, and other dementias and conditions. We have been administering the test using a digitizing ballpoint pen that reports its position with considerable spatial and temporal precision, making available far more detailed data about the subject’s performance. Using pen stroke data from these drawings categorized by our software, we designed and computed a large collection of features, then explored the tradeoffs in performance and interpretability in classifiers built using a number of different subsets of these features and a variety of different machine learning techniques. We used traditional machine learning methods to build prediction models that achieve high accuracy. We operationalized widely used manual scoring systems so that we could use them as benchmarks for our models. We worked with clinicians to define guidelines for model interpretability, and constructed sparse linear models and rule lists designed to be as easy to use as scoring systems currently used by clinicians, but more accurate. While our models will require additional testing for validation, they offer the possibility of substantial improvement in detecting cognitive impairment earlier than currently possible, a development with considerable potential impact in practice.

ObjectiveTo determine whether a digital clock-drawing test, DCTclock, improves upon standard cognitive assessments for discriminating diagnostic groups and for detecting biomarker evidence of amyloid and tau pathology in clinically normal older adults (CN).MethodsParticipants from the Harvard Aging Brain Study and the PET laboratory at Massachusetts General Hospital were recruited to undergo the DCTclock, standard neuropsychological assessments including the Preclinical Alzheimer Cognitive Composite (PACC), and amyloid/tau PET imaging. Receiver operating curve analyses were used to assess diagnostic and biomarker discriminability. Logistic regression and partial correlations were used to assess DCTclock performance in relation to PACC and PET biomarkers.ResultsA total of 300 participants were studied. Among the 264 CN participants, 143 had amyloid and tau PET imaging (Clinical Dementia Rating [CDR] 0, Mini-Mental State Examination [MMSE] 28.9 ± 1.2). An additional 36 participants with a diagnosis of mild cognitive impairment or early Alzheimer dementia (CDR 0.5, MMSE 25.2 ± 3.9) were added to assess diagnostic discriminability. DCTclock showed excellent discrimination between diagnostic groups (area under the receiver operating characteristic curve 0.86). Among CN participants with biomarkers, the DCTclock summary score and spatial reasoning subscores were associated with greater amyloid and tau burden and showed better discrimination (Cohen d = 0.76) between Aβ± groups than the PACC (d = 0.30).ConclusionDCTclock discriminates between diagnostic groups and improves upon traditional cognitive tests for detecting biomarkers of amyloid and tau pathology in CN older adults. The validation of such digitized measures has the potential of providing an efficient tool for detecting early cognitive changes along the AD trajectory.Classification of EvidenceThis study provides Class II evidence that DCTclock results were associated with amyloid and tau burden in CN older adults.