Background and Aims
The aim was to review existing models of post-stroke cognitive impairment (CI) disease progression, to inform development of the StrokeCog epidemiological model of post-stroke CI.
Method
Pubmed, EMBASE and the NHS Economic Evaluations Database were searched for relevant articles published in English since 2000. Included articles described an epidemiological or economic model tracking disease progression of CI and/or stroke over time, e.g., Markov or microsimulation models. Titles and abstracts were screened by one reviewer, followed by full text of potentially relevant articles. Data was abstracted from eligible articles using a standard form. Models were evaluated against pre-specified criteria for the StrokeCog model, agreed by expert consensus.
Results
5,723 abstracts were screened, followed by full text review of 398 articles. There were 72 eligible articles for CI and 54 for stroke (two articles modelled stroke and CI). The dominant model type in both reviews was a Markov cohort-level state transition model. Models of CI assume continuous disease progression, whereas models of stroke assume stability in the absence of stroke recurrence. 
Conclusion
To date, epidemiological modelling of post-stroke cognitive impairment has been extremely limited. It requires a synthesis of two existing approaches - modelling the ongoing risk of transition to a worse cognitive state, and the increased risk of deterioration associated with a recurrent stroke.

POST-STROKE COGNITIVE IMPAIRMENT: A SYSTEMATIC REVIEW OF DISEASE PROGRESSION MODELS