Definition
Estimated fatigue risk level of crew assignments that are HOS-compliant, using validated fatigue modeling (SAFTE-FAST, Biomathematical Fatigue Model, or equivalent) to measure the gap between "legally allowed to work" and "physiologically likely to be alert."
The Measurement Gap
The Measurement Gap
Crew schedulers verify HOS compliance before making assignments. What they don't verify is whether an HOS-compliant assignment is actually safe from a fatigue science perspective. The Hours of Service regulations set maximum on-duty time and minimum rest periods, but they don't account for circadian rhythm effects, cumulative sleep debt, or the difference between "off duty" and "sleeping." A crew member who is legal under HOS but coming off a string of early-morning starts with disrupted sleep can be cognitively impaired to a level equivalent to 0.06% BAC.
This gap is not theoretical. The NTSB has cited schedule-induced fatigue as a contributing factor in multiple rail accidents where all crew HOS requirements were met. Fatigue modeling tools (SAFTE-FAST, Biomathematical Fatigue Models) exist and are used in aviation, but rail schedulers don't incorporate them as constraints.
Decision Supported:
Is the crew scheduler optimizing assignments to the HOS boundary in ways that are legal but unsafe? Should the scheduler incorporate fatigue-model constraints beyond HOS compliance? FMBCI provides the population-level risk indicator to drive scheduler policy changes.
Formula
Plain Language: For each crew assignment, run the crew member's recent work/rest history through a validated fatigue model. Compare the model's predicted effectiveness score at the assignment's critical operating periods (start, midpoint, end) against a fatigue threshold. Count assignments where the crew member is HOS-legal but fatigue-model-impaired.
Notation:
FMBCI = Assignments_legal_but_fatigued / Total_assignments
Variables:
| Variable | Definition | Unit | Source | Classification |
|---|---|---|---|---|
| Assignments_legal_but_fatigued | Crew assignments where HOS compliance = true AND fatigue_model_effectiveness < fatigue_threshold at any critical operating point | Count | Crew system + fatigue model | DERIVED |
| Total_assignments | All crew assignments in the measurement period | Count | Crew management system | NATIVE |
| fatigue_threshold | Effectiveness score below which cognitive performance is materially impaired (recommended: 77 on SAFTE-FAST scale, equivalent to 0.06% BAC cognitive impairment) | Score | Fatigue model standard | REFERENCE |
Sign Convention: 0% = all legal assignments are also fatigue-safe (ideal). Higher = more legal-but-impaired assignments. Track by time of day, assignment pattern (early-morning starts, back-to-back nights), and corridor.
Edge Cases:
- Fatigue model input availability: SAFTE-FAST requires sleep/wake history, which may not be captured. Alternative: use HOS rest periods as a proxy for sleep, with a discount factor (not all off-duty time is sleep time). This reduces accuracy but maintains computability.
- Individual variation: fatigue models predict population averages. Individual tolerance varies. Use the model as a population risk indicator, not an individual diagnostic.
- Minimum sample: 200 assignments per measurement period.
Proposed formula, pending validation pilots.