Clinical Report: Predictors in Post-Stroke Delirium Prediction Models
Overview
This systematic review identifies key predictors of post-stroke delirium (PSD) from existing models and assesses their methodological quality. The pooled area under the curve (AUC) indicates moderate to good discrimination, but high risk of bias raises concerns about the reliability of these models.
Background
Post-stroke delirium is a significant complication that can adversely affect recovery and outcomes in stroke patients. Understanding the predictors of PSD is crucial for early identification and management, which can improve patient care. This review synthesizes existing prediction models to inform clinical practice and future research.
Data Highlights
Study Count
Models
Sample Size Range
Pooled AUC
16
24
100 - 14,475
0.83 (95% CI: 0.81–0.85)
Key Findings
Age, NIHSS score, neutrophil-to-lymphocyte ratio, visual impairment, and infection are significant predictors of PSD.
The pooled AUC for model discrimination was 0.83, indicating moderate to good performance.
All included studies exhibited a high overall risk of bias, primarily due to methodological shortcomings.
Calibration was acceptable in six studies, but clinical utility was rarely evaluated.
Future studies should standardize PSD diagnostic criteria and utilize robust validation strategies.
Clinical Implications
Highlight the necessity for clinicians to assess the applicability of findings critically.
Conclusion
This review underscores the importance of identifying predictors of PSD while highlighting the need for improved methodological quality in future studies. Reliable prediction models are crucial for enhancing patient outcomes in stroke care.
