Objective:

This study aims to identify unique clinical profiles in patients with lacunar stroke in the National Inpatient Sample (NIS) by leveraging machine learning-based clustering algorithms.

Background:

Not Applicable

Design/Methods:

The 2015-2019 NIS was queried using ICD-10 PCS coding to identify patients with lacunar stroke. A machine learning clustering algorithm evaluated the population based on 50 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 12,083 patients were included in this study. Composite DBI and CHI scoring determined the optimal number of clusters to be seven, with sizes ranging from 58-9437 patients. Mortality ranged from 1.07 % in Cluster 1 to 27.93% in Cluster 7. Clusters 3,4,5,6 and 7 each displayed significantly higher rates of mortality [OR Range 2.38 - 35.83, p<0.001] relative to Cluster 1. Individual cluster profiles are visualized in a heatmap. Cluster 1 had the greatest prevalence of hyperlipidemia amongst groups. Cluster 4 had the greatest prevalence of sepsis, arrhythmia, and aspiration pneumonia. Risk of non-routine discharge was highest in Cluster 7 [OR 11.21, p<0.001]. Kruskal-Wallis H-Testing and post-hoc pairwise comparison of length of stay distributions showed significant (p<0.001) differences between all clusters except 2 and 3, 5 and 6 and 6 and 7 with the greatest test statistics occurring when comparing group 1 to all other groups.

Conclusions:

Clustering analysis of patients with lacunar stroke identified 7 distinct groups. This clustering approach enables a nuanced understanding of comorbidity interactions, further informing clinical decision-making.