Using A.I. to ID Epilepsy Surgery Candidates
The decision to refer a patient for an epilepsy surgery consult is rarely obvious, and often happens later than it has to in the course of treatment. Researchers at Cincinnati Children’s Hospital Medical Center have developed a novel algorithm that uses artificial intelligence and machine learning to help accelerate the process.
The algorithm runs on a software program that is embedded in the hospital’s electronic health record (EHR). Using natural language processing techniques, it analyzes previous provider notes for each patient who has an upcoming appointment in one of the hospital’s epilepsy clinics. The algorithm considers the words, tone and themes in the notes, and uses a scoring system to identify patients who meet neurosurgical criteria.
Real-time decision support The researchers are conducting a prospective study to evaluate the algorithm’s effectiveness in flagging patients who are possible surgical candidates, and test the best way to notify providers, whether by email or an alert in the EHR. “This project is one of the first in the field to study the integration of natural language processing recommendations into clinical care,” says Judith Dexheimer, PhD, principal investigator of the study. “The algorithm is designed to provide real-time clinical decision support, and ultimately reduce the time to initial surgery evaluation for patients with intractable epilepsy.”
Epilepsy is one of the leading neurological disorders in the United States, affecting more than 479,000 children. Approximately 30 percent of epileptic patients have poor seizure control despite antiepileptic medications, and are potential candidates for neurosurgical intervention. While relevant guidelines exist for making a referral for a neurosurgical intervention, there is no standardized way of doing so.
The study is a collaborative effort of the John Pestian Lab at Cincinnati Children’s and the hospital’s divisions of neurology and neurosurgery. The Pestian Lab, which focuses on biomedical informatics, is conducting a number of artificial intelligence studies to decrease the amount of time it takes for patients to be referred for epilepsy surgery. The national average in pediatrics is 10 years.
An efficient analysis “Given the vast amount of note-based data available in the electronic health record, it is challenging for providers to efficiently retain and process all the pertinent patient information over a long period of time,” says Dexheimer. “The epilepsy surgery decision support tool can summarize notes much more efficiently than a person can, and prompt providers to take a fresh look at a patient’s needs. We will be evaluating how this algorithm affects referral patterns and, over time, whether it leads to an increase in epilepsy surgeries.”
Other study collaborators at Cincinnati Children’s include Hans Greiner, MD, co-director of the hospital’s Epilepsy Surgery Program, Katherine Holland-Bouley, MD, PhD, director of neurophysiology, and Francesco Mangano, DO, division chief of pediatric neurosurgery. The study ends this fall and Dexheimer hopes that, if successful, the decision support tool will be expanded to adult data and implemented in other institutions around the United States. Meanwhile, her research team is seeking collaborators who may be interested in testing the tool at their own institutions. To learn more, email Judith.Dexheimer@cchmc.org.
