Individualized Performance Prediction During Total Sleep Deprivation: Accounting for Trait Vulnerability to Sleep Loss
ARMY MEDICAL RESEARCH AND MATERIEL COMMAND FORT DETRICK MD TELEMEDICINE AND ADVANCED TECH RESEARCH CENTER
Pagination or Media Count:
Individual differences in vulnerability to sleep loss can be considerable, and thus, recent efforts have focused on developing individualized models for predicting the effects of sleep loss on performance. Individualized models constructed using a Bayesian formulation, which combines an individuals available performance data with a priori performance predictions from a group-average model, typically need at least 40 h of individual data before showing significant improvement over the group-average model predictions. Here, we improve upon the basic Bayesian formulation for developing individualized models by observing that individuals may be classified into three sleep-loss phenotypes resilient, average and vulnerable. For each phenotype, we developed a phenotypespecific group-average model and used these models to identify each individuals phenotype. We then used the phenotypespecific models within the Bayesian formulation to make individualized predictions. Results on psychomotor vigilance test data from 48 individuals indicated that, on average, -85 of individual phenotypes were accurately identified within 30 h of wakefulness. The percentage improvement of the proposed approach in 10-h-ahead predictions was 16 for resilient subjects and 6 for vulnerable subjects. The trade-off for these improvements was a slight decrease in prediction accuracy for average subjects.
- Stress Physiology
- Statistics and Probability