Precise Prediction and Treatment of Seizures After Intracranial Hemorrhage

Seizures are a common and morbid complication of intracranial hemorrhage, leading to brain herniation, worse patient outcome, and death. While a few risk factors for seizures have been described, the ability to predict seizures is still crude. The inaccuracy of seizure prediction leads to imprecise administration of prophylactic antiseizure medications. Prophylactic antiseizure medications are intended to prevent seizures, reduce complications, and improve patient outcomes. Unfortunately, antiseizure medications have been independently associated with more complications, worse patient outcomes, and worse health-related quality of life (HRQoL), particularly cognitive function HRQoL. Better methods are needed to predict precisely which patients are likely to have seizures after intracranial hemorrhage to prevent them, and, further, to determine which patients are likely to benefit from prophylactic antiseizure medications. We will continue a successful line of research. At the time we began to study this topic, prophylactic phenytoin was recommended by guidelines. After publications implicated phenytoin with more complications and worse patient outcomes, guidelines were changed to discourage the use of prophylactic phenytoin, and clinicians broadly switched from phenytoin to levetiracetam. We recently reported that prophylactic levetiracetam is independently associated with worse cognitive function HRQoL in the 40% of patients who receive it, underscoring that current practice may lead to inadvertent harm, an untenable status quo. The effects of seizures on HRQoL are worse than prophylactic antiseizure medication. Preventing seizures by precise administration of prophylactic antiseizure medication would be helpful. This proposal has two major aims that will improve patient outcomes after intracranial hemorrhage. First, we will build upon our previous work to derive and validate a multi-dimensional model for predicting seizures after intracranial hemorrhage from electroencephalography and imaging data to identify the patients most likely to benefit from prophylactic seizure medication. A prospective database with recording of seizures and patient outcomes provides preliminary data. Future data will be electronically abstracted from a health care system with a single electronic health record using automated techniques from electroencephalography reports, raw electroencephalography data, and neuroimaging source data. Then, we will determine the effect of prophylactic seizure medication on patient HRQoL at high risk for seizures, and use other machine learning techniques to determine which patients are most likely to benefit from prophylactic antiseizure medication. At the conclusion of this proposal, we will deliver a model to predict patients most likely to have seizures, and determine which patients are likely to have higher HRQoL as a results of prophylactic seizure medications, leading to targeted treatment and non-treatment to maximize patient HRQoL.