Shift-work impairment is usually treated as a sleep-quantity problem. It is better described as a misalignment problem: the internal clock is in one place, the schedule demands another, and performance falls in the gap. Misalignment can be quantified — and once quantified, it can be managed.
The gold standard for circadian phase is dim-light melatonin onset — a saliva or plasma series collected under controlled light. It is accurate and completely impractical outside a lab.
Core body temperature is the next best anchor, and its minimum sits in a stable relationship to melatonin onset. Core temperature is also awkward to measure. But distal skin temperature — fingers, toes — swings with peripheral vasodilation, which is itself under circadian control, and inverts the core signal with a predictable lag.
A finger-worn thermistor sampling continuously therefore carries a usable phase signal, provided you can reject the confounds. That rejection — ambient compensation, activity gating, off-wrist detection, illness flags — is where most of our engineering has gone.
A phase estimate that is right on average but silent about its own uncertainty is worse than no estimate, because people act on it during a shift.
Figures are pilot results, not registrational evidence. Study protocols and analysis code are available to collaborating institutions on request — see Publications.
Full inversion is optimal for a fixed night rota and harmful for a fast rotation. The crossover point is not well characterised in free-living populations.
Light prescriptions work when followed. Our own retention data is the honest constraint on how much benefit we can claim at a population level.
If phase is known for a whole team, rosters could be built to minimise aggregate impairment. Whether that survives real staffing constraints is an open question.
We supply devices and raw data to labs working on circadian measurement.