| Autor: |
Ghassemi, Mohammad M., Amorim, Edilberto, Al Hanai, Tuka, Lee, Jong W., Herman, Susan T., Sivaraju, Adithya, Gaspard, Nicolas, Hirsch, Lawrence J., Scirica, Benjamin M., Biswal, Siddharth, Moura, Valdery, Cash, Sydney S., Brown, Emery N., Mark, Roger G., Westover, M. Brandon |
| Jazyk: |
angličtina |
| Rok vydání: |
2019 |
| Předmět: |
|
| Zdroj: |
Crit Care Med |
| Popis: |
OBJECTIVE: Electroencephalogram (EEG) features predict neurological recovery following cardiac arrest. Recent work has shown that prognostic implications of some key EEG features change over time. We explore whether time dependence exists for an expanded selection of quantitative EEG (QEEG) features and whether accounting for this time-dependence enables better prognostic predictions. DESIGN: Retrospective. SETTING: Intensive care units at four academic medical centers in the U.S. PATIENTS: Comatose patients with acute hypoxic-ischemic encephalopathy. INTERVENTIONS: None. MEASUREMENTS: We analyzed 12,397 hours of EEG from 438 subjects. From the EEG, we extracted 52 features that quantify signal complexity, category, and connectivity. We modeled associations between dichotomized neurological outcome (good vs. poor) and QEEG features in 12-hour intervals using sequential logistic regression with Elastic-Net regularization. We compared a predictive model utilizing time-varying features to a model using time-invariant features and to models based on two prior published approaches. Models were evaluated for their ability to predict binary outcomes using area under the receiver operator curve (AUC), model calibration (how closely the predicted probability of good outcomes matches the observed proportion of good outcomes), and sensitivity at several common specificity thresholds of interest. MAIN RESULTS: A model utilizing time-dependent features outperformed (AUC = 0.83 ± 0.08) one trained with time-invariant features (0.79 ± 0.07, p |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
