Systematic review of post-stroke delirium prediction models finds moderate discriminative ability

AIM: To systematically review published studies on the post stroke delirium risk prediction models; and to provide the evidence for developing and updating the clinically available prediction models. DESIGN: Systematic review. DATA SOURCES: Systematically searched studies on 10 databases, which were conducted from inception to 9 January 2025. The studies of post-stroke delirium risk prediction models were included. METHODS: Extracted the data from the selected studies. The Prediction Model Risk of Bias Assessment Tool checklist was used to evaluate the risk of bias of the models. The meta-analysis of model performance and common predictors was performed by Revman 5.4 and Medcalc. RESULTS: A total of 12 studies were included, and 21 risk prediction models for post-stroke delirium were constructed. The combined effect size of area under the receiver operating characteristic curve was 0.84. All studies were found to have a high risk of bias and good applicability. Meta-analysis showed: National Institutes of Health Stroke Scale score, age, neutrophil-to-lymphocyte ratio, neglect, visual impairment and atrial fibrillation were independent predictors of post-stroke delirium. CONCLUSION: The included studies all found to have a high risk of bias; future studies should focus on adopting more scientifically rigorous study designs and following the standardised reporting guidelines to enhance extrapolation and facilitate its clinical application. IMPLICATIONS FOR THE PROFESSION: This review may promote clinical healthcare workers to develop and update clinically available prediction models, thereby establishing risk prediction models with strong clinical utility. IMPACT: This study presents the first systematic evaluation of delirium risk prediction models in stroke patients, thereby facilitating the choice, use and develop of the clinical usable post stroke delirium risk prediction models. REPORTING METHOD: This review adhered to the PRISMA guidelines. PATIENT OR PUBLIC CONTRIBUTION: No patient or public contribution. REVIEW REGISTRATION: RD42024620360 (PROSPERO According to JAN Guidelines).