Tiffany Ong, BS
Wake Forest School of Medicine
Winston Salem, North Carolina
Disclosure information not submitted.
Priscilla Zameza, NA
Medical Student
Wake Forest School of Medicine, United States
Disclosure information not submitted.
Stacey Wolfe, MD
Associate Professor, Neurosurgery
Wake Forest School of Medicine, United States
Disclosure information not submitted.
Umit Topaloglu, PhD
Associate Professor, Cancer Biology
Wake Forest School of Medicine, United States
Disclosure information not submitted.
Pam Duncan, PhD
Professor, Neurology
Wake Forest School of Medicine, United States
Disclosure information not submitted.
Mohd Anwar, PhD
Professor, Computer Sciences
North Carolina A&T State University, United States
Disclosure information not submitted.
Raymond Samuel, MD, PhD
Professor, Department of Biology
North Carolina A&T State University, United States
Disclosure information not submitted.
Bhavana Budigi, MD
Fellow, Neuroradiology
Wake Forest School of Medicine, United States
Disclosure information not submitted.
Chris Lack, MD
Assistant Professor, Neuroradiology
Wake Forest School of Medicine, United States
Disclosure information not submitted.
Aarti Sarwal, MD, FAAN, FNCS,FCCM
Medical Director, Neurocritical Care
Wake Forest Baptist Health Center
Winston-Salem, North Carolina
Disclosure information not submitted.
Title: Distance Travelled To Tertiary Care As Prognostic Indicator In Intracerebral Hemorrhage Outcomes
Introduction: Intracranial hemorrhage (ICH) has high morbidity and mortality, disproportionately affecting rural patients despite adjusting for comorbidities. Inter-hospital transfers for rural patients cause delays in access to specialized care and are associated with adverse outcomes. Published prognostic tools lack distance as factor hence we explored training of three machine learning models to predict 30-day mortality, modified Rankin scale on discharge and discharge disposition in ICH patients using distance from home to tertiary care.
Methods: Preprocessing functions and ML models were imported from the Python 3.8 library scikit-learn. All categorical variables were one-hot-encoded; ordinal variables were integer encoded. Three machine learning models were trained to predict three labels: 30 Day Mortality (Alive, Dead/Hospice), modified Rankin Score upon discharge (7 classes), and Discharge Disposition (Home/Inpatient Rehabilitation Facility, Hospice/Acute Care Facility Skilled Nursing Facility/Other Health Care Facility/Expired /Long Term Care Hospital). Data was split 60/40 for training and testing sets respectively. Mean and standard deviation of F1 score, recall, and precision were calculated over 10 trials. Feature importance was determined using permutation feature importance.
Results: The dataset contained 138 patients admitted in calendar year 2019 with 13 useable features: 8 categorical features, 4 numeric features, and 1 ordinal feature. For all of these models, the five most important features were GCS on Admission, ICH Score, age, smoking status, and maximum distance travelled to tertiary care facility. The Multinomial Naive Bayes model showed F1 score of 0.8129 for 30-day mortality. Random forest performed lower but better than chance for mRS at discharge, F1 score 0.68 and discharge disposition 0.34. Varying classification of discharge disposition did not improve
Conclusion: Distance travelled by an ICH patient from home to tertiary medical center was shown to be indeterminant in patient’s outcome in our predictive model. Ongoing study including rural urban status, time traveled to specialized care from point of origin as well as payer status are being evaluated in a larger sample to create a prognostic model accounting for social determinants affecting ICH outcomes.