GRAPHICAL MODELLING OF THE INDEPENDENT ASSOCIATIONS BETWEEN DOMAINS SPECIFIC COGNITIVE IMPAIRMENTS IN SUBACUTE STROKE PATIENTS
- Owen Williams (United Kingdom)
Abstract
Background And Aims
Cognitive impairments are highly prevalent after stroke, often with multiple domain impairments presenting simultaneously. It remains to be understood how deficits in different cognitive functions interact with each other. For example, are impairments in domain general functions such as executive functions more likely to impact the overall cognitive profile compared to domain specific impairments in e.g. spatial attention. A Graphical modelling approach established a network of associations, with metrics of centrality denoting the relative strength of connections between specific domain tests.
Methods
Participants were sub-acute stroke survivors recruited for three separate studies, two in the UK and one in Italy. In total 1035 participants (452 Female, mean age = 70.8 years) were included in analyses. Cognitive domains were assessed using 11 subtests from the Oxford Cognitive Screen and binarized into impaired vs. unimpaired. The Bootnet R package was used to apply an isofit model the binary data. Bootstrapping was used to assess graph stability and compare the expected influence of each test.
Results
Figure 1 shows the resulting cognitive impairment network. The mixed trails test (a measure of executive function) had the highest expected influence. It was significantly stronger than the expected influence for tests of episodic memory, praxis, semantic knowledge, and visual spatial neglect. However, the expected influence of the trails test was not significantly higher than that for language processing, number processing, or orientation. 
Conclusions
These results provide evidence that impairments in executive functions and language have a large negative impact on other cognitive domains.