We submitted an entry to the 2023 PhysioNet Challenge (team WesternUni). The Challenge dataset contains EEG recordings, clinical information, and functional outcomes of cardiac arrest patients taken within 72 hours following ICU admission. To predict patients' functional outcome, we trained an ensemble of extreme gradient-boosted (XGB) algorithms on several EEG and patient features. Predictions were made and optimized at three steps to maximize the probability of correct prediction. Before model training, the raw EEG recordings were band-pass filtered [0.1-30Hz], motion-corrected, downsampled to 100Hz, then combined into 75-second epochs. We then computed features on every epoch, including power spectral density (from 1 – 30 Hz), complexity (e.g., Lempel-Ziv, fractal dimension, entropy), spectral coherence and weighted phase lag index (in delta, theta, and alpha ranges) and basic statistical properties of the signal (e.g., mean, variance), each of which was then averaged across epochs. On the first step of model training, XGB algorithms were trained on each feature individually to predict the functional outcome associated with the recording. Next, each feature from the previous step was averaged within each patient and then fit with an XGB algorithm. Patient-specific clinical information (e.g., age, ROSC) was also fit with an XGB algorithm at this step. Last, an XGB algorithm was trained on all the outcome predictions made in previous steps to produce the final outcome prediction. Hyperparameter tuning (via the tree-structured Parzen Estimators Approach) occurred at each step and feature, with the objective of optimizing balanced accuracy on 20 stratified cross-validations of the training data. This stepwise approach could successfully predict outcomes across time points on the challenge's validation dataset, producing initial Challenge Scores of 0.40 (12 hrs), 0.54 (24 hrs), 0.58 (48 hrs) and 0.58 (72 hrs). These results show the potential effectiveness of this technique for prognosticating critically-ill patients following cardiac arrest.