Abstract Assoziation studies revolutionized genomics by rigorously screening large feature sets against health outcomes. Digital medicine now produces similarly high-dimensional, longitudinal sensor data from wearables, smartphones, and connected environments. We propose Sensor-Wide Association Studies (SWAS): structured, feature-wide, hypothesis-generating scans of a pre-specified library of sensor-derived features against one or more pre-defined clinical phenotypes, with transparent feature documentation, appropriate longitudinal modeling, and principled control of multiplicity. This perspective outlines minimal standards, common failure modes, and ethical considerations to help SWAS become a reproducible foundation for digital epidemiology and personalized medicine.
Steckhan et al. (Sat,) studied this question.