Related Experiment Videos
Machine Learning Characterization of Readmissions After Chronic Subdural Hematoma Hospitalizations
Huanwen Chen1, Marco Colasurdo2, Matthew K McIntyre3
1Neurology, University of Maryland Medical Center, Baltimore.
Neurology
|August 7, 2026
Summary
Readmissions after chronic or subacute subdural hematoma (cSDH/sSDH) are common and diverse, often unrelated to surgical recurrence. Machine learning identified distinct patient groups, necessitating tailored care strategies beyond surgical focus.
Area of Science:
- Neurosurgery
- Geriatrics
- Data Science
Background:
- Readmission after chronic or subacute subdural hematoma (cSDH/sSDH) hospitalization is frequent.
- Current focus is mainly on surgical recurrence, neglecting the full spectrum of readmission causes and patient heterogeneity.
Purpose of the Study:
- To characterize the incidence, diversity, and outcomes of 90-day readmissions after cSDH/sSDH hospitalization.
- To identify patient phenotypes with distinct readmission risk profiles using machine learning.
Main Methods:
- Retrospective cohort study using the Nationwide Readmissions Database (2016-2022).
- Included adults nonelectively hospitalized for cSDH/sSDH.
- Machine learning-based phenotyping (Shapley Additive Explanations, K-means clustering) identified patient subgroups.
Main Results:
- 29.0% of patients were readmitted within 90 days, with diverse causes.
- Non-SDH readmissions (e.g., infection) had higher mortality and disability rates than surgical recurrence.
- Five distinct patient clusters with unique readmission patterns were identified.
Conclusions:
- cSDH/sSDH readmissions are diverse and predominantly nonsurgical, with significant clinical impact.
- Machine learning revealed substantial patient heterogeneity, requiring tailored therapeutic and care strategies.
- Clinical trials and postdischarge care models should expand beyond surgical recurrence to address diverse patient subgroups.