Neural activity is known to correlate with decrements in task performance as individuals enter the state of mental fatigue which might lead to lowered productivity and increased safety risks. Incorporating a passive brain computer interface (BCI) technique that detects changes in subject's neural activity and predicts the behavioral performance when the subject is underperforming might be a promising approach to reduce human error in real-world situations. Here, we developed a reliable model using EEG power spectrum to estimate time-on-task performance in a psychomotor vigilance test (PVT) which can fit across individuals. High correlation between the estimated and actual reaction time was achieved. Hence, our results illustrate the feasibility for modeling time-on-task decrements in performance among different individuals from their brainwave activity, with potential applications in several domains, including traffic and industrial safety.