Game-Based Assessment of Cognitive Status in Clinical and Non-Clinical Samples

In this dissertation I validated a serious game for cognitive assessment that was originally developed in my Masters thesis (Tong, 2014). I used a whack-a-mole game that requires response inhibition, by adding a distractor character that should not be hit. I evaluated the game as a potential screening tool for the clinically relevant condition of delirium, and I examined how game performance was correlated with the Mini-Mental State Examination (MMSE), a standard assessment of cognitive ability in the elderly. Patients 65 years and older presenting at the Sunnybrook emergency department participated in three clinical studies and response time (RT) on the game was found to be significantly correlated with MMSE scores, and sufficiently predictive of whether the person had delirium as determined by the Confusion Assessment Method. Comparing game performance at 8-hour intervals, I found that game RT showed good test-retest reliability in the second clinical study. In the third clinical study, I found that false alarm rate provided a useful supplement to median correct RT in predicting clinical measures. Following the clinical studies, I evaluated the game with healthy adults in a pilot study to determine how the duration of character appearance affected whether the task was speeded or not. I found that a character appearance duration of around 800 msec marked the transition from a speeded to an unspeeded task. I then carried out a final experiment with healthy adults where the two appearance durations used were 800 msec (borderline speeded) and 1200 msec (unspeeded). The experiment assessed game performance for eight combinations of character appearance duration and two other dichotomous factors, and assessed performance on four tasks that measure different aspects of inhibition ability (Stroop task, Flanker task, anti-saccade task, Go/No-Go discrimination task). False alarms increased significantly in the condition where targets were least frequent and distractors were visually similar to the target. Game performance was strongly correlated with performance on the Go/No-Go discrimination task but not with the other three inhibition tasks used. The results of this thesis point the way towards the development of serious games that can supplement current clinical assessments.Ph.D