EEG-Based Risk Score May Predict Likelihood of Seizures in Hospitalized Patients

A risk score based on EEG variables and seizure history may quickly and accurately aid clinical judgment of seizure risk in patients who are critically ill, according to research published online ahead of print October 9 in JAMA Neurology.

Aaron F. Struck, MD

Studies have detected a high incidence of subclinical seizures in patients with sepsis, traumatic brain injury, and other conditions, but no simple, validated method exists to assess a patient’s seizure risk using a combination of factors. Aaron F. Struck, MD, Assistant Professor of Neurology at the University of Wisconsin in Madison, and colleagues sought to create a simple scoring system associated with the probability of seizures in patients with acute illness.

Data From Three Centers

The investigators created the scoring system using data from a multicenter prospective database. The database included clinical and electrographic variables from patients who had more than six hours of uninterrupted EEG recordings for clinical indications other than epilepsy monitoring unit admissions. The continuous EEG sessions were recorded at Emory University Hospital in Atlanta, Brigham and Women’s Hospital in Boston, and Yale New Haven Hospital in Connecticut. In all, 5,427 continuous EEGs were performed on 4,772 patients (49.9% men; median age, 61).

To build their scoring model, the investigators used a machine-learning method, Risk-Calibrated Supersparse Linear Integer Model (RiskSLIM), that considered 24 candidate variables. The researchers used cross-validation to validate the model’s accuracy and risk calibration.

The final model, which the researchers called 2HELPS2B, had an area under the curve of 0.819 and an average calibration error of 2.7%. A patient’s score ranges from 0–7 and is based on the following six variables:

• brief (ictal) rhythmic discharges (B[I]RDS) (2 points)
• presence of lateralized periodic discharges, lateralized rhythmic delta activity, or bilateral independent periodic discharges (1 point)
• prior seizure (1 point)
• sporadic epileptiform discharges (1 point)
• frequency greater than 2.0 Hz for any periodic or rhythmic pattern (1 point)
• presence of “plus” features (ie, superimposed, rhythmic, sharp, or fast activity) (1 point).

The seizure risk associated with each score from 0–5 was 5%, 12%, 27%, 50%, 73%, and 88%, respectively. For a score of 6 or 7, the probable seizure risk was greater than 95%.

The large sample size and use of data from multiple centers are among the study’s strengths. Limitations of the study include that duration of EEG was not included in the database, and that no sessions of less than six hours were included in the study.

“The 2HELPS2B score is an easy-to-use tool to augment clinical judgment of the risk for seizures in individual patients,” Dr. Struck and colleagues said. “The simplicity of the system allows for easy integration into clinical workflow. With increasing familiarity, 2HELPS2B will improve communication between EEG interpreters and clinicians through the use of a quickly comprehensible single number to describe seizure risk for patients on continuous EEG.”

Questions Remain

The investigators designed “a simple scale with good accuracy, which can be easily used by clinicians to estimate seizure risk in their patients,” said Barry M. Czeisler, MD, Assistant Professor of Neurology and Neurosurgery at New York University School of Medicine, and Jan Claassen, MD, PhD, Associate Professor of Neurology at Columbia University College of Physicians and Surgeons in New York, in an accompanying editorial. Still, the score is unvalidated for prediction based on less than six hours of EEG recordings and should be validated in prospective studies, they said.

A seizure risk scale has the potential to inform clinical practice. “Certain patients may not need to stay on continuous EEG for a long time if their seizure risk is low,” but an acceptable level of risk remains unclear, said Drs. Czeisler and Claassen. In addition, the score potentially could guide which patterns of features warrant more aggressive treatment, such as with additional antiepileptic medication. “Appropriate risk stratification using 2HELPS2B may allow us to answer these questions adequately in the near future,” they said. “The development of adequate measurement tools is often necessary to appropriately study a condition, which in turn may allow for future optimization of treatment algorithms.”

—Jake Remaly

Suggested Reading

Czeisler BM, Claassen J. A novel clinical score to assess seizure risk. JAMA Neurology. 2017 Oct 9 [Epub ahead of print].

Struck AF, Ustun B, Rodriguez Ruiz A, et al. Association of an electroencephalography-based risk score with seizure probability in hospitalized patients. JAMA Neurology. 2017 Oct 9 [Epub ahead of print].