Title

Prediction model for risk of developing seizures after cardiac arrest

Document Type

Poster

Publication Date

4-30-2020

Institution/Department

Maine Medical Center, Medical Education, Maine Medical Center Research Institute, Critical Care, Neurology & Neuroscience

MeSH Headings

Heart Arrest, Seizures

Abstract

Background / Purpose Seizures are common following resuscitation from cardiac arrest and are often the result of hypoxic ischemic encephalopathy; many are only detectable by electroencephalography (EEG). The presence of seizures is associated with worse outcomes, but the effect of early identification and treatment may be protective. Monitoring for seizures requires intensive resources which are not available at all centers. The ability to risk-stratify patients at high risk would be helpful to allocate limited resources such as continuous EEG monitoring and increased frequency of EEG review. Our aim with this study was to create a practical model for seizure prediction in this population.

Methods / Approach We used existing data from Maine Medical Center patients enrolled in the International Cardiac Arrest Registry (INTCAR) from 2008-2019. All were treated under protocols and received targeted temperature management and standardized intensive care; most underwent continuous EEG. We included those >18 years, hospitalized following resuscitation from cardiac arrest. Differences in characteristics of patients with and without seizures were compared. A logistic regression model was created, using patient and arrest characteristics available at the time of hospital admission and with P<0.2 or clinically important to predict the risk for an individual patient to develop seizures. 95% confidence intervals were used, and a P-value ?0.05 was considered significant. All analyses were conducted in the statistical software R (version 1.2.5033).

Results We included 794 patients. Seizures were identified in 95 (12%) patients. Average age was 59 years (SD=16), number of patients with a shockable rhythm was 382 (48%) and 120 (15%) patients were previously healthy. In the final model, patients who were previously healthy (OR=0.43, 0.19-0.87), and patients with witnessed arrest (OR= 0.57, 0.34-0.98) were less likely to have seizures, while PEA as initial rhythm (OR=1.74, 1.02-2.96), not following commands due to sedation at hospital admission (OR=1.90, 1.17-3.03) and normal ECG at admission (OR=2.11, 0.83-4.91) were more likely to have seizures. The c-statistic of the model was 0.664 (0.609-0.719). Hosmer and Lemeshow goodness of fit test of the model showed good discrimination (p=0.62).

Conclusions In patients with cardiac arrest, several factors on hospital admission are associated with the development of seizures. Although the prediction of seizures is difficult for an individual patient, these factors may be useful in discriminating between high and low risk patients for the purpose of EEG monitoring.

Comments

2020 Costas T. Lambrew Research Retreat, abstract only

Full text is not available here. Please contact your library for more information.

Share

COinS