Machine Learning Predicts Cerebral Vasospasm in Subarachnoid Hemorrhage Patients: Multi-Center Validation

Friday, May 3, 2024

3:10 PM – 3:14 PM CT

Location: S401ab

Presenting Author(s)

David Zarrin, MSE

Medical Student
UCLA Medicine
Bethesda, MD, US

Co-Author(s)

GC

Geoffrey Colby, MD, PhD

Associate Professor, Program Director
University of California Los Angeles
Los Angeles, California, United States
RF

Robert E. Freundlich, MD, MS, MSCI

Associate Professor of Anesthesiology and Biomedical Informatics
Vanderbilt University Medical Center, United States
EG

Eilon Gabel, MD, MHA

Associate Professor of Anesthesiology
University of California, Los Angeles Department of Anesthesia and Perioperative Medicine, United States
BG

Bilwaj Gaonkar, PhD

Neurosurgical Researcher
University of California, Los Angeles Department of Neurosurgery, United States
LM

Luke Macyszyn, MD

Neurosurgeon
University of California, Los Angeles Department of Neurosurgery
Marina del Rey, California, United States
KM

Karen McCarthy, MS, EdD

Assistant Director, Vanderbilt Anesthesiology & Perioperative Informatics Research
Vanderbilt University Medical Center, United States
Abhinav Suri, MPH (he/him/his)

Medical Student
David Geffen School of Medicine at UCLA
Los Angeles, California, United States
BW

Bayard Wilson, MD

Neurosurgery Resident
University of California, Los Angeles Department of Neurosurgery, United States

Disclosure(s):

David Zarrin, MSE: No financial relationships to disclose

Introduction: Cerebral vasospasm (CV) is a feared complication occurring in 20-40% of patients following subarachnoid hemorrhage (SAH) and is known to contribute to delayed cerebral ischemia. It is standard practice to admit SAH patients to intensive care for an extended period of vigilant, resource-intensive, clinical monitoring. We now present multi-center validation of our machine learning approach for predicting CV requiring verapamil (CVRV).

Methods: SAH patients admitted to UCLA from 2013-2022 and a validation cohort from Vanderbilt University Medical Center (VUMC) from 2018-2023 were included. For each patient, 172 unique intensive care unit (ICU) variables were extracted through the primary endpoint, namely intraoperative verapamil or not. At each institution, a light gradient boosting machine (LightGBM) was trained using five-fold cross validation to predict the primary endpoint at various timepoints during hospital admission. Receiver-operator curves (ROC) and precision-recall (PR) curves were generated.

Results: A total of 1,750 UCLA patients were included, 125 receiving verapamil. LightGBM achieved an area under the ROC (AUC) of 0.88 over one week in advance when predicting verapamil or not, and successfully ruled out 8% of non-verapamil patients with zero false negatives. Minimum leukocyte count, maximum platelet count, and maximum intracranial pressure were the variables with highest predictive accuracy. Our models predicted “no CVRV” vs “CVRV within three days” vs “CVRV after three days” with AUCs=0.88, 0.83, and 0.88, respectively. From VUMC, 1,654 patients were included, 75 receiving verapamil. VUMC predictive models performed very similarly to UCLA models, averaging 0.01 AUC points lower.

Conclusion : We present an accurate (AUC=0.88) and early (>1 week prior) predictor of cerebral vasospasm requiring verapamil using machine learning over two large cohorts of subarachnoid hemorrhage patients at separate institutions. This represents a significant step towards optimized clinical management and improved resource allocation in the intensive care setting of subarachnoid hemorrhage patients.