Objectives: This study examined the influence of a wrist-worn heart rate drowsiness detection device on heavy vehicle driver safety and sleep and its ability to predict driving events under naturalistic conditions. Design: Prospective, non-randomized trial. Setting: Naturalistic driving in Malaysia. Participants: Heavy vehicle drivers in Malaysia were assigned to the Device (n = 25) or Control condition (n = 34). Intervention: Both conditions were monitored for driving events at work over 4-weeks in Phase 1, and 12-weeks in Phase 2. In Phase 1, the Device condition wore the device operated in the silent mode (i.e., no drowsiness alerts) to examine the accuracy of the device in predicting driving events. In Phase 2, the Device condition wore the device in the active mode to examine if drowsiness alerts from the device influenced the rate of driving events (compared to Phase 1). Measurements: All participants were monitored for harsh braking and harsh acceleration driving events and self-reported sleep duration and sleepiness daily. Results: There was a significant decrease in the rate of harsh braking events (Rate ratio = 0.48, p < 0.05) and a fall in subjective sleepiness (p < 0.05) when the device was operated in the active mode (compared to the silent mode). The device predicted when no driving events were occurring (specificity=98.81%), but had low accuracy in detecting when a driving event did occur (sensitivity=6.25%). Conclusions: Including drowsiness detection devices in fatigue management programs appears to alter driver behaviour, improving safety despite the modest accuracy. Longer term studies are required to determine if this change is sustained.