P072 Predicting vigilance in a shift-work context with under-mattress sensor data from the preceding sleep

Abstract Introduction Performance impairments are risky in many industries, including healthcare and defence. Predicting when more likely to occur is critical reduce costly workplace accidents errors. This study created models for predicting vigilance during simulated night shift-work from under-mattress sleep sensor data. Methods The parent compared two conditions phase delaying individuals adjust shift-work. Participants (N=11) attended the laboratory eight days per condition, one month apart. After a baseline (10PM-7AM), participants remained awake 27hrs, then slept 10AM-7PM next four with metrics recorded by an (the Withings Sleep Analyzer). At night, completed shifts, six psychomotor tasks (PVTs) shift (48 resulting datapoints participant). current predicted PVT performance preceding daytime based on 27 variables using machine learning (Extra trees) models. Data were randomly split into 67% subset model training variable reduction, 33% testing fit. Results 12 retained following feature reduction. final demonstrated good fit reaction time <500ms (R² = 0.79, RMSE 20.4ms), reciprocal 0.70), number of lapses 0.69). Discussion These preliminary findings comparable fatigue prediction models, supporting that can be unobtrusively collected Further research will confirm whether these may assist safer delegation work self-management shift-workers.