Clustering and Subgroup Analysis of Subarachnoid Hemorrhage and Obesity in 3,500 Patients Using Machine Learning

Introduction: Subarachnoid hemorrhage (SAH) is a highly acute condition requiring prompt management to prevent mortality and morbidity. This study employs machine learning-based clustering algorithms to identify unique clinical profiles in patients with SAH and obesity, providing new insights into the stratification of risk factors among this population.

Methods: The 2015-2019 National Inpatient Sample was queried using ICD-10 CM/PCS coding to identify patients with Obesity and SAH. Machine learning clustering analysis evaluated the population based on 47 comorbidities, complications and clinical covariates. Optimal number of clusters was determined using the Davies-Bouldin Index (DBI) and Calinski-Harabasz Index (CHI). Between-cluster multivariate logistic regression analysis was performed to assess risk of mortality and non-routine discharge. Kruskal-Wallis H-Testing was performed to assess variance in length-of-stay between clusters. Statistical analysis was performed using Python.

Results: A total of 3,543 patients were grouped into five clusters based on composite DBI and CHI scoring. Cluster sizes ranged from 117-2,468 patients. Mortality ranged from 12.60-46.29% between Clusters 1 and 5. Cluster 5 had the greatest prevalence of acute kidney failure (AKF), pneumonia, sepsis, and mechanical ventilation status. Cluster 5 was also associated with a 5-fold increase in mortality risk [OR 5.96, p< 0.001] relative to Cluster 1. Kruskall-Wallis H-testing of length-of-stay distributions showed significant differences (p < 0.001) among clusters. Post-hoc pairwise comparison testing showed the greatest differences in medians occurring when comparing Cluster 1 to all others.

Conclusion : Clustering analysis of patients with subarachnoid hemorrhage and obesity identified 5 distinct groups with unique comorbidity profiles. The cluster with the highest prevalence of AKF, pneumonia, sepsis, and mechanical ventilation status was also associated with the greatest risk of mortality. This machine learning clustering approach enables a nuanced understanding of previously imperceptible comorbidity interactions, providing a basis for more personalized clinical decision-making and treatment interventions.