Comparing clock tests for dementia screening: Naïve judgments vs formal systems - What is optimal?

Background: Clock drawing tests (CDTs) vary in format, scoring, and complexity. Herein, we compared the dementia screening performance of seven CDT scoring systems and the judgements of untrained raters. Methods: 80 clock drawings by subjects of known dementia status were selected, 20 from each of four categories (Consortium to Establish a Registry for Alzheimer's disease [CERAD] defined normal, mild, moderate, and severe abnormality). An expert rater scored all clocks using published criteria for seven systems. Additionally, 20 naïve raters judged clocks as either normal or abnormal, without formal instructions. Clocks were then classified by drawers' dementia status for comparison of dementia detection across systems. Results: Naïve and formal CDT systems showed 90-100% agreement in CERAD normal, moderate and severe categories, but poor agreement (mean = 39%) for mildly impaired clocks. When CDT systems were compared for accurate dementia classification, the Mendez and CERAD systems correctly identified the greatest proportion of subjects (84-85%), and Wolf-Klein the smallest (58%). The better systems correctly identified > 70% of mildly demented individuals (CDR = 1). In contrast, medical records from patients' personal physicians correctly identified only 24% of the mildly demented. Strikingly, naïve raters' CDT judgements were as effective as five of the seven CDT systems in dementia identification. Conclusions: While the Mendez system was the most accurate overall, it was not significantly better than CERAD, which had simpler scoring rules. Untrained raters discriminated normal from abnormal clocks with acceptable accuracy for community screening purposes. Results suggest that, if used, most CDT systems would improve personal physicians' dementia recognition in difficult to detect mildly demented subjects.