Automatic assessment of upper limb function during play of the action video game, circus challenge: validity and sensitivity to change

Abstract

We present an evaluation of the cross-sectional and longitudinal validity (sensitivity to change) of a new algorithm to assess upper limb function generated automatically during play of a bespoke, professionally-written action video game (Circus Challenge Assessment Game, CCAG). The subjects were 33 patients with hemiplegia after stroke (aged 33-81 years), who played the CCAG on 8 separate occasions over a 12 week period; 19 in the chronic phase after stroke, with stable function; 14 in the acute phase during recovery of function. A convergent construct validation process was used by correlating the scores of the CCAG with a validated, clinically assessed reference, the Chedoke Arm and Hand Assessment Inventory (CAHAI). Cross- sectional validity was demonstrated using the between-subjects correlation coefficient (r = .998, p-value <; .001); longitudinal validity by the within-subjects correlation coefficient (overall, r = .54, p-value <; .001; chronic, r = .33, p-value <; .001; acute, r = .63, p - value <; .001). There was no difference between the CCAG and CAHAI in the classification of pa- tients to acute and chronic groups (comparing ROC curves, p - value = .50), demonstrating similar sensitivity to change. This is the first time a serious game can achieve automatic in- game assessment to a clinical standard, solely using low cost, commodity hardware and professionally written action video games and demonstrates the potential for remote monitoring of patients during home based rehabilitation programs.