Unsupervised clustering based on a graph attention network reveals high-risk subgroups among ICU dementia patients for enhanced mortality prediction.

backgroundDementia presents a significant global health burden, with an increasing prevalence and considerable economic implications. In the intensive care unit (ICU), the number of dementia patients has steadily increased, highlighting the need for more accurate risk stratification. Traditional methods of patient classification often assume homogeneity, limiting their ability to identify subgroups at varying levels of mortality risk. This study explored the potential of graph attention networks (GATs) in addressing this gap, aiming to better characterize high-risk dementia subgroups in the ICU and improve clinical outcomes.

methodsThis study utilized anonymized patient data from the publicly available MIMIC-IV and MIMIC-III datasets. Dementia patients were identified using ICD-9 and ICD-10 codes, with exclusions based on specific admission criteria, resulting in the inclusion of 7904 patients. A comprehensive set of demographics, comorbidities, ICU severity scores, and laboratory parameters was selected. Three unsupervised clustering methods were employed: a graph autoencoder (GAE), a graph attention autoencoder (GAT), and K-means combined with UMAP-based dimensionality reduction. The optimal number of clusters (k = 3) was determined by maximizing the silhouette scores and Calinski-Harabasz indices. The GAT method quantitatively outperformed both the GAE and K-means methods, achieving superior cluster separation (MIMIC-IV: highest silhouette score: 0.090; Calinski-Harabasz: 345.501; Davies-Bouldin: 2.30; MIMIC-III: highest silhouette score: 0.210; Calinski-Harabasz: 380.620; Davies-Bouldin: 1.850). Three distinct subgroups were identified: Cluster 1 represented the highest-risk phenotype, characterized by advanced age, coagulopathy, anemia, and organ dysfunction, with a strikingly elevated mortality risk (HR = 21.58, p < 0.001). Compared with the unstratified cohort, the machine learning models trained on the MIMIC-IV and validated on the MIMIC-III cohort showed improved predictive performance (AUC up to 0.85 within clusters). SHAP analysis revealed cluster-specific drivers, including a prolonged length of stay, renal impairment, and inflammatory markers. Visualization of the subgroups was achieved via sunburst plots, radar charts, and boxplots.

resultsThis study included 7904 patients, who were divided into two groups: the nonoccurrence group (n = 5,038) and the occurrence group (n = 2,866). Baseline characteristics, including demographic, clinical, and laboratory factors, significantly differed between the groups. Unsupervised clustering methods identified distinct patient subgroups, with the GAT algorithm outperforming K-means and GCN in clustering both cohorts. Cox regression revealed that the GAT identified high-risk groups, with mortality significantly higher in Cluster 1 (HR = 21.58, P < 0.001). Cluster analysis revealed that Cluster 1 had the highest mortality risk and was characterized by advanced age, severe renal dysfunction, and poor clinical outcomes. To validate the superiority of the GAT-based unsupervised clustering approach, we applied various machine learning algorithms, which demonstrated enhanced predictive performance across the identified subgroups. The feature distributions and interrelationships were visualized using radar plots, boxplots, and SHAP analysis. SHAP further revealed distinct risk profiles: Cluster 0 was associated with an increased risk of thrombotic events and acute kidney injury; Cluster 1 reflected a high metabolic stress state indicative of early systemic inflammatory response syndrome (SIRS) or sepsis; and Cluster 2 was primarily marked by severe renal failure.

conclusionsThis study demonstrates the effectiveness of GAT-based clustering for identifying high-risk subgroups among dementia patients in the ICU. By providing more granular risk stratification, these models can support clinical decision-making and improve patient management. These findings underscore the potential of advanced machine learning techniques in enhancing ICU care for dementia patients, with future research focusing on temporal validation and integration into clinical workflows to optimize outcomes.