Prediction models for aspiration risk in stroke patients: a systematic review

Background: Stroke is a leading cause of disability and death, with post-stroke dysphagia significantly increasing aspiration risk, leading to complications, such as aspiration pneumonia and higher mortality. Various prediction models for aspiration exist, but their clinical utility is limited by methodological heterogeneity. Aim: This review aimed to evaluate the performance and applicability of these models for stroke patients, informing future model optimization and clinical use. Study design: A comprehensive search was conducted across nine electronic databases until 20 January 2025. Studies on the development or validation of aspiration risk prediction models in adult stroke patients were included. Data extraction followed the CHARMS checklist, and bias risk and applicability were assessed using the PROBAST tool. The review was prospectively registered with PROSPERO (CRD420251007112). Results: Eighteen model development studies were included. Most demonstrated good discrimination (AUC/C-index: 0.756-0.955), with 16 showing good applicability. However, all studies had a high risk of bias, mainly due to retrospective designs, small sample sizes (events per variable < 20), inadequate missing data handling, univariate variable selection, and limited external validation. Key predictors included age, NIHSS score, Kubota Water Swallowing Test, history of aspiration, and Glasgow Coma Scale score. Conclusion: Aspiration risk prediction models for stroke patients show promising predictive performance but are limited by methodological bias and heterogeneity. Future research should prioritize rigorous reporting and multi-center, large-sample external validation to improve model robustness and clinical applicability. Systematic review registration: PROSPERO CRD420251007112, URL: https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD420251007112.