Online monitoring technology for deep phenotyping of cognitive impairment after stroke

Abstract Background Despite the high prevalence of disabling post-stroke cognitive sequalae, these impairments are often underdiagnosed and rarely monitored longitudinally. Provision of unsupervised remote online cognitive technology would provide a scalable solution to this problem. However, despite recent advances, such technology is currently lacking, with existing tools either not meeting the scalability challenge or not optimised for specific applications in post-stroke cognitive impairment. To address this gap, we designed and developed a comprehensive online battery highly optimised for detecting cognitive impairments in stroke survivors. Method The technology is optimised to allow both diagnosis and monitoring of post-stroke deficits, and for remote unsupervised administration. Participants performed 22 computerised tasks, and answered neuropsychiatric questionnaires and patient reported outcomes. 90 stroke survivors (Mean age = 62.1 years; 68% and 32% in the acute and subacute/chronic phase after stroke respectively) and over 6,000 age-matched healthy older adults were recruited. Patient outcome measures were derived from Bayesian Regression modelling of the large normative sample and validated against standard clinical scales. Results Our online technology has greater sensitivity to post-stroke cognitive impairment than pen-and-paper tests such as the MOCA (mean sensitivity 81.75% and 52.25% respectively, P<0.001). Further, our outcomes show a stronger correlation with post-stroke quality of life (r(78)=0.51, R2=0.26, P<0.001) when compared to MOCA, which only explains half of this variance (r(78)=0.38, R2=0.14, P< 0.001). An additional set of experiments confirm that the online tasks yield highly reliable outcomes, with consistent performance observed across supervised versus unsupervised settings, and minimal learning effects across multiple timepoints. Conclusion The current online cognitive monitoring technology is feasible, sensitive, and reliable when assessing patients with stroke. The technology offers an economical and scalable method for assessing post-stroke cognition in the clinical setting and sensitively monitoring cognitive outcomes in clinical trials for stroke.