Introduction: Cognitive impairment (CI) is a frequent consequence of stroke and is associated with increased costs and reduced quality of life. However, its inclusion in model-based economic evaluation for stroke is limited. Objective: To identify, review, and critically appraise current models of stroke for use in economic evaluation, and to identify applicability to modeling poststroke CI. Methods: PubMed, EMBASE, and the NHS Economic Evaluations Database (NHS EED) were systematically searched for papers published from January 2008 to August 2018. Studies that described the development or design of a model of stroke progression intended for use in economic evaluation were included. Abstracts were screened, followed by full text review of potentially relevant articles. Models that included CI were retained for data extraction, and among the remainder, models that included both stroke recurrence and disability were also retained. Relevance and potential for adaptation for modeling CI were assessed using a standard questionnaire. Results: Forty modeling studies were identified and categorized into 4 groups: Markov disability/recurrence (k = 29); CI (k = 2); discrete event simulation (k = 4), and other (k = 5). Only 2 modeling studies included CI as an outcome, and both focused on narrow populations at risk of intracranial aneurysm. None of the models allowed for disease progression in the absence of a stroke recurrence. None of the included studies carried out any sensitivity analysis in relation to model design or structure. Conclusions: Current stroke models used in economic evaluation are not adequate to model poststroke CI or dementia, and will require adaptation to be used for this purpose.There were substantial commonalities in model design and structure, with 4 groupings identified: Markov disability and recurrence model, Markov cognition model, DES, and other.
Group 1: Markov Disability and Recurrence Model (k = 29)This group included cohort level Markov models. These models tracked cohorts of patients through 2 or more disability states, recurrent stroke, and death. Disability states were generally defined using the mRS (27/29 studies) [14]. This model structure did not allow any disease progression (deterioration or improvement) in the absence of stroke recurrence, apart from 2 models which allowed for progression during a short "rehabilitation" phase [15,16].Deterministic and probabilistic sensitivity analyses were used to handle uncertainty in parameter estimates. However, there was limited or no sensitivity analysis in relation to model structure or key underlying assumptions regarding disease progression. Only 3/29 reported validation against observational data [15,17,18]. dementia. We aim to synthesize an approach that models stroke disease progression as an ongoing risk of stroke recurrence, with the CI/dementia approach which models a continuous risk of cognitive decline (see http://www.strokecog.ie/ for more details).
ConclusionThere is limited diversity in model design and structure in model-bas...