Prediction Models for Post-Stroke Cognitive Impairment: A Systematic Review and Meta-Analysis.

Abstract

Background: Stroke is one of the most serious illnesses worldwide and is the primary cause of acquired disability among adults. Post-stroke cognitive impairment (PSCI) is a complication of stroke that significantly impacts patients' daily activities and social functions. Therefore, developing a risk prediction model for PSCI is essential for identifying and preventing disease progression.

Objectives: This study systematically reviewed and analyzed PSCI prediction models, identifying the associated risk factors.

Methods: We systematically retrieved literature from PubMed, Cochrane Library, Embase, and other sources. Two researchers independently extracted the literature and assessed the risk of bias using the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS) checklist and The Prediction Model Risk of Bias Assessment Tool (PROBAST).

Results: A total of 20 articles describe the PSCI prediction model, with an incidence rate ranging from 8% to 75%. The area under the receiver operating characteristic curve (AUC) value for the development models ranged from 0.66 to 0.969, while the validation models ranged from 0.763 to 0.893. Age, diabetes, hypersensitive C-reactive protein (hs-CRP), hypertension, and homocysteine (hcy) were identified as the strongest predictors.

Conclusion: In this systematic review, several PSCI prediction models demonstrate promising prediction performance, although they often lack external validation and exhibit high heterogeneity in some predictive factors. Therefore, we recommend that medical practitioners utilize a comprehensive set of predictive factors to screen for high-risk PSCI patients. Furthermore, future research should prioritize refining and validating existing models by incorporating novel variables and methodologies.