Repeated computerized cognitive testing: Performance shifts and test-retest reliability in healthy older adults

Introduction: Repeated cognitive assessment is frequently required to monitor changes in cognitive functioning in older adults, but studies of repeated computerized testing in this population are scarce. To provide new insight into retest effects this study examined within- and between-day performance shifts and test-retest reliability among healthy older adults for test scores from a computerized cognitive battery. Method: Thirty older men (65-71 years) completed the battery six times. Testing occurred twice on each of three testing days, separated by 1 week. Results: Reaction times (RTs) on tasks intended to measure inhibition (Anti), response switching (Pro/Anti), selective attention (Simon and Flanker), and working memory (2-back) typically showed practice effects, which were most prominent between the first two time points. In most cases, these RTs showed moderate to good test-retest reliability (intraclass correlation coefficient, ICC, range = .34 to .93) with lower reliability between the first two time points. Two-back accuracy rates showed similar results. In contrast, RTs on a basic visuomotor task (Pro) and on compatible trials of the Simon task showed increases at later time points, presumably because of boredom, but demonstrated mostly moderate to good reliability (ICC range = .49 to .83). Scoring metrics from a computerized version of the Corsi Block-Tapping task (intended to measure short-term and working memory) and cost scores (performance differences between two related conditions/tasks) intended to isolate specific cognitive functions tended to show poor reliability (ICC range = -.23 to .62). Conclusions: Most of the RT tasks investigated showed suitability for use in repeated testing among older adults, although longer familiarization periods appear to be warranted in many cases to reduce practice effects and improve initial reliability. However, poor reliability indicated that scoring metrics from the computerized Corsi Block-Tapping task and cost scores are unsuitable for repeated testing.