Association of an Electroencephalography-Based Risk Score With Seizure Probability in Hospitalized Patients.

Area of Science:

• Neuroscience
• Critical Care Medicine
• Medical Informatics

Background:

• Continuous electroencephalography (EEG) is increasingly used in critical care.
• A validated method to assess seizure risk in these patients is lacking.

Purpose of the Study:

• Develop a simple scoring system to estimate seizure probability in acutely ill patients.
• Utilize EEG and clinical history seizure risk factors for the scoring system.

Main Methods:

• A prospective, multicenter study involving 5427 continuous EEG sessions.
• Developed a scoring model using the RiskSLIM machine learning method.
• Validated the model's accuracy and risk calibration via cross-validation.

Main Results:

• The 2HELPS2B model demonstrated strong performance with an AUC of 0.819 and average calibration error of 2.7%.
• The model incorporates 6 key variables, including specific EEG patterns and prior seizure history.
• Probable seizure risk ranges from 5% to over 95% based on the score.

Conclusions:

• The 2HELPS2B model provides a rapid and accurate tool for clinical judgment of seizure risk.
• This scoring system can aid clinicians in managing critically ill patients with potential seizure activity.