Cluster Analysis of Patients With Acute Ischemic Stroke: Identifying Characteristics of Long Hospital Stays in a Comprehensive Hospital

Background

Acute ischemic stroke (AIS) is a leading cause of mortality and disability worldwide, imposing a significant burden on patients and healthcare systems. Length of stay (LOS) is a critical metric for assessing hospital resource utilization and patient prognosis. Identifying characteristics of AIS patients with prolonged LOS is essential for optimizing resource allocation and improving patient management. This study aimed to use cluster analysis to profile AIS patients with long LOS in a comprehensive hospital and explore differences in characteristics across gender and age subgroups.

Methods

This single-center, retrospective cohort study included 664 patients admitted with AIS to Peking Union Medical College Hospital from June 2012 to September 2021. Data collected included demographics, admission NIHSS scores, stroke risk factors, etiologies, and diagnostic workups. Patients were clustered using the K-prototype method, a machine learning technique for subclassifying complex data, to differentiate between patients with long and short LOS. Statistical tests were used to identify significant differences between the clusters.

Results

Cluster analysis revealed that patients with longer LOS had a higher proportion of females (42.9% vs. 24.7%, p < 0.001) and were generally younger (52.3 vs. 65.4 years, p < 0.001). This group exhibited lower proportions of TOAST type 1 strokes (17.7% vs. 70.4%, p < 0.001), higher levels of hsCRP and D-dimer, and no significant difference in acute phase NIHSS scores. Notably, in-hospital strokes and admissions to non-neurological departments were more frequent in the long LOS group. Subgroup analysis by gender and age revealed that younger males and females shared similar characteristics with the overall long LOS group, including a higher incidence of non-neurological department admissions and higher D-dimer levels.

Conclusions

This study highlights the heterogeneity of AIS and the importance of etiological identification, particularly in younger and female patients. Our findings suggest that traditional factors like NIHSS scores may not fully capture the complexity of factors influencing LOS in these groups. Improved cross-departmental collaboration is crucial for better management of AIS patients.