Combined clustering and association-rule analysis for hospital costs and length of stay in spontaneous intracerebral hemorrhage hospitalizations.

Introduction: Intracerebral hemorrhage (ICH) imposes a substantial burden on inpatient resources, yet hospital costs and length of stay (LOS) are often evaluated as separate endpoints. This study aimed to identify interpretable clinical-administrative strata and recurrent high-utilization care patterns among spontaneous ICH hospitalizations. Methods: This single-center retrospective cohort study included 1,851 spontaneous ICH hospitalizations from 5 November 2022 to 31 October 2024. Clustering variables were restricted to age, admission source, diagnosis-derived hemorrhage-location category, primary ventricular hemorrhage/intraventricular extension, hypertension-coded status, diabetes, and chronic kidney disease/renal failure. Cost-composition ratios, total cost, LOS, procedures, discharge disposition, departments, and acute in-hospital complications were excluded from cluster formation and reserved for descriptive analyses, adjusted modeling, or association-rule mining. Gower dissimilarities with k-medoids clustering were evaluated across K = 2-8 using elbow analysis, average silhouette coefficient, permutation-based gap statistic, subsampling stability, cluster size, clinical interpretability, and parsimony. Results: The Discussion: This pathway-oriented framework organized ICH hospitalizations into interpretable baseline clinical-administrative strata and identified recurrent comorbidity/procedure/complication patterns associated with high utilization. These findings may support risk-aware benchmarking and resource management under diagnosis-related group/diagnosis-intervention packet payment reform, although validation using granular clinical severity data is required.