Prediction of Pilot Performance During Startle Events Based on Neuropsychophysiological Features of Stress Resilience and Cognitive Task Scores

The disruptive impact of the startle effect on pilots can lead to potentially fatal outcomes, highlighting the importance of identifying novel predictors of pilot performance during unexpected and unpredictable startle events. Accordingly, this study investigates the potential of optimally selected stress resilience features and cognitive task scores to predict pilot performance during startle events. Seventeen military pilot cadets participated in a two-phase experiment conducted over two weeks. In Phase 1, context-free stimulation paradigms including neuropsychophysiological measurements were administered. In Phase 2, specific flight simulator sessions with startle events were undertaken. Optimal neuropsychophysiological stress resilience features and cognitive task scores from Phase 1 were selected as predictors of pilot simulator performance in Phase 2 using multiple regression analysis (MRA) within a leave-one-out cross-validation (LOOCV) procedure. The obtained MRA models demonstrated significant predictive validity across left-out participants, as indicated by a Pearson correlation coefficient of 0.7039 (p = 0.0034) between the actual and predicted simulator performance, with a mean absolute error (MAE) of 2.5457° versus a dummy median regressor MAE of 3.4881°. The optimally predictive features/scores from Phase 1 were related to startle reflex habituation, multitasking ability, heart rate recovery after threat, and resting respiratory sinus arrhythmia. These findings may improve the accuracy of predicting pilot performance during startle events, with implications for pilot selection and training.