Citation

Raslear TG, Coplen M. Aviat. Space Environ. Med. 2004; 75(3 Suppl): A168-72.

Affiliation

Office of Research and Development, Federal Railroad Administration, Washington, DC 20590, USA. Thomas.Raslear@fra.dot.gov

Copyright

(Copyright © 2004, Aerospace Medical Association)

DOI

unavailable

PMID

15018281

Abstract

Human fatigue models are increasingly being used in a variety of industrial settings, both civilian and military. Current uses include education, awareness, and analysis of individual or group work schedules. Perhaps the ultimate and potentially most beneficial use of human fatigue models is to diagnose if an individual is sufficiently rested to perform a period of duty safely or effectively. When used in this way, two important questions should be asked: 1) What is the accuracy of the diagnosis for duty-specific performance in this application; and 2) What decision threshold is appropriate for this application (i.e., how "fatigued" does an individual have to be to be considered "not safe"). In the simplest situation, a diagnostic fatigue test must distinguish between two states: "fatigued" and "not fatigued," and the diagnostic decisions are "safe" (or "effective") and "not safe" (or "not effective"). The resulting four decision outcomes include diagnostic errors because diagnostic tests are not perfectly accurate. Moreover, since all outcomes have costs and benefits associated with them that differ between applications, the choice of a decision criterion is extremely important. Signal Detection Theory (SDT) has demonstrated usefulness in measuring the accuracy of diagnostic tests and optimizing diagnostic decisions. This paper describes how SDT can be applied to foster the development of fatigue models as practical diagnostic and decision-making tools. By clarifying the difference between accuracy (or sensitivity) and decision criterion (or bias) in the use of fatigue models as diagnostic and decision-making tools, the SDT framework focuses on such critical issues as duty-specific performance, variability (model and performance), and model sensitivity, efficacy, and utility. As fatigue models become increasingly used in a variety of different applications, it is important that end-users understand the interplay of these factors for their particular application.

Language: en