Beyond Feature Importance: Feature Interactions in Predicting Post-Stroke Rigidity with Graph Explainable AI

TL;DR

The paper tackles predicting post-stroke rigidity by foregrounding feature interactions through graph-based explainable AI, using 519K HCUP-US stroke hospitalizations to compare traditional models (Logistic Regression, XGBoost, Transformer) with graph-based approaches (Graphormer, GATv2). Graph-based models demonstrate AUROC around 0.75, outperforming baselines and providing intrinsic explanations that capture interactions among features, particularly during-hospital clinical assessments such as NIHSS and APR-DRG severity/risk mortality. The study reveals consistent top predictors, uncovers interaction pathways, and offers personalized, patient-level explanations that can guide early identification and tailored rehabilitation; despite limitations, the work demonstrates the practical potential of graph-based XAI for stroke prognosis. Overall, the approach advances interpretability and predictive power by explicitly modeling feature interactions, enabling clinicians to understand both population-level and individual patient risk profiles for post-stroke rigidity.

Abstract

This study addresses the challenge of predicting post-stroke rigidity by emphasizing feature interactions through graph-based explainable AI. Post-stroke rigidity, characterized by increased muscle tone and stiffness, significantly affects survivors' mobility and quality of life. Despite its prevalence, early prediction remains limited, delaying intervention. We analyze 519K stroke hospitalization records from the Healthcare Cost and Utilization Project dataset, where 43% of patients exhibited rigidity. We compare traditional approaches such as Logistic Regression, XGBoost, and Transformer with graph-based models like Graphormer and Graph Attention Network. These graph models inherently capture feature interactions and incorporate intrinsic or post-hoc explainability. Our results show that graph-based methods outperform others (AUROC 0.75), identifying key predictors such as NIH Stroke Scale and APR-DRG mortality risk scores. They also uncover interactions missed by conventional models. This research provides a novel application of graph-based XAI in stroke prognosis, with potential to guide early identification and personalized rehabilitation strategies.

Figures (2)

• Figure 1: Diagram of the Beyond Feature Importance: Feature Interactions in Predicting Post-Stroke Rigidity with Graph Explainable AI.
• Figure 2: Personalized explanations from the Graph Learning Model using GATv2.