Before applying a wearable alcohol study to an individual case, identify the device model, the reference measure and whether the authors tested their rule in an independent group.
Follow the threshold from development to testing
Ash and colleagues developed classification rules for a wrist-worn BACtrack Skyn using one cohort and tested them in another. Specificity fell to 70% when the derived threshold was applied to the second cohort. The study also reported weaker classification performance in an analysis of devices further from their shipment date.
Match the study to the question
Check whether the published metrics address individual samples, events or days. Then compare the device version and field conditions to those in the study. A daily classification metric based on mobile diary reports answers a different question from the accuracy of a single ankle-monitor alert.
Source
Ash et al., 2022 study of wearable alcohol monitor performance
Leave a Reply